Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Energy Bytes

Energy Bytes artwork

EP 20: Peter Harding from Kelvin Inc.

Nov 29, 20231 hr 15 minAudio

Peter Harding, founder of Kelvin discuss how his company is transforming the oil and gas industry with cutting-edge autonomous operations. Peter shares his journey from his beginnings in California to becoming a key innovator in energy tech startups. The conversation delves into the technical complexities of autonomous oilfields, the role of AI in enhancing efficiency, and the challenges of integrating advanced technology into the sector. Peter emphasizes the importance of human expertise working alongside AI, aiming to augment rather than replace human labor, and provides insights into the future of AI-powered solutions in oilfield operations. Join the conversation shaping the future of energy. Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward. Apply today at collide.io Click here to view the episode transcript. 0:00 - Podcast Introduction 0:50 - Peter Harding Background 5:40 - Kelvin Origin Story 9:32 - Kelvin's Current Role in AI 10:17 - AI Adoption Strategies 14:05 - AI as a Force Multiplier 15:31 - Knowledge Institutionalization 16:50 - AI Implementation Ease 19:39 - Industry Adoption Speed 21:55 - Empathy in AI Sales 24:50 - Identifying Customer Pain Points 26:45 - Transition to AI Software 30:37 - Data Quality Significance 38:50 - System Level AI Control 40:37 - Causal Inference in AI 44:12 - AI as a Complementary Tool 46:50 - Hybrid AI Models 48:54 - AI Edge Case Challenges 50:08 - Combining Human Expertise with AI 50:35 - AI in Reservoir Impact Examples 55:41 - Data Science Scalability 59:47 - Data Science Simplification 1:01:04 - Energy Sector Software Learning Curve 1:03:03 - Comprehensive vs Specialized Solutions 1:05:08 - Specialization in Energy Tech 1:06:20 - Quickfire Questions Segment 1:10:02 - Energy Tech Career Advice 1:14:04 - Texas Ice Storm Discussion 1:14:44 - Connecting with Peter Harding https://twitter.com/collide_ai https://www.tiktok.com/@collide.io https://www.instagram.com/collide.io/ https://www.facebook.com/collide.io/ https://www.linkedin.com/company/collideai

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show.

Welcome back everybody to uh another episode of Energy Bites. Brad Dad is here today. Uh we had Fuse earlier this week, so thanks everybody that came out. I'm actually repping a uh honeycomb shirt. Thanks Oleg. Shout out to Oleg for uh for giving us some stuff. I've actually decked out my uh my cup as well with a bunch of new uh honeycomb uh stickers. They're pretty good. Oleg, I know you were behind those, so keep doing them. They're legit. here with uh Bobby today. You know, I've got one for you. It's in the chat. I just forgot to bring it inside. And today we've got the founder of Kelvin, Peter Harding. Hey, nice to meet you guys. Thanks for thanks for coming on, man. Absolutely. Happy to be here. Yeah, I'm glad we could uh catch you while you were in town. Thatdeed that works out uh works out well. Well, let's jump into it. So, um you're the founder CEO of Kelvin currently. kind of walk us through, you know, where you from, where you grew up, how you got kind of into energy, and then to a founder of a energy tech startup. Totally. Yeah. So, uh, again, thanks for the time today. So, uh, born and raised in the Bay Area in California and was really active sort of as a kid in a lot of kind of different engineering endeavors and got excited about building stuff. Fast forward, uh, I was thinking about the one of the big catalyst events for me was actually the the Persian Gulf War around an awareness around energy and and a recognition that I didn't really understand why we had a whole bunch of our troops over there and and was trying to sort of rationalize that as a kid. And that was that was interesting in the sense also because my dad's a two-time, you know, Vietnam vet, like a couple tours and and had really like, you know, gone out and and made that that sacrifice. And I was trying to understand sort of his effort and commitment versus the what we were why we were out there sort of in, you know, in the Gulf War. And that sort of energy dependence as a kid really resonated with me. And so I started to think about what what types of technologies and other things in the future are going to change that. So fast forward um I became aware of a lot of the kind of interesting work that was starting to happen around directional drilling and fracking kind of going back you know this is you know pushing 20 20 plus years ago and as a kid from California that's not a it's not necessarily top of mind. Um but for me it was just it was one of these like both of those technologies were so transformative that um but not necessarily appreciated in parts of the Silicon Valley. Uh still aren't still aren't right. And and I and I say that with like I I have the same sort of appreciation that I think many folks in Texas do for frankly how hard it is to to to do that and to make those systems. It's really mind-blowing when you think about what miles underneath, miles out. Exactly. 14,000 foot lateral. Nobody talks about it. No. And and and so when I started to think about it, I'm like, "Wow, that is that's really interesting. What types of things can I bring to potentially be a you know, sort of as a as a as an entrepreneur and a person who's interested in building things? What types of technology can I bring to bear to potentially aid and support some of these things?" Because I do believe that where we are today at 13 million barrels per day as of you know back I think that I saw that statistic right which is sort of the you know that was pre- pandemic levels we're back to that that's a function of what we've done on sort of the work in the US right and that's amazing yeah it's funny when you think about that and you look at the uh the uh peak oil decline curves and stuff and you're like it goes like this and then the shale boom kicks off and it just comes right back up. And I think that's a just sidebar, but I I think that's the thing that somebody mentioned this actually at Fuse. Um you know, when they when people account for technology and they forecast stuff, they don't forecast you can't, right? You can't forecast new things that haven't been invented yet or that haven't been proven yet or things like that. But the weird thing about commodities is the higher the commodities price, the more potential there, right? $14 and MCF gas and all of a sudden hey you're like exactly [laughter] well I was with sort of fast forward till we can talk about but just on along those lines um a couple weeks ago I was back at the the climate impact conference one of our investors is OGCI or now climate impact and Vicky Halib was talking about some of the work that Oxie is doing right and you know today they expect you know traditional expectations or we get about 10% you know out of recovery factor, what our recovery factor is. And you look at some of the potential in terms of new ELR techniques and they're talking about like 19%. Yeah, it's insane. I mean, even like I think that was something that almost an arbitrage that they're thinking with Exxon because Exxon's over here pumping that they can they're going to 2x the recovery factor. Well, if you just 2x your perian footprint and you can 2x that now you got 4x potential, you like. So, yeah. So, I I get excited about step function changes that that technology can bring and and I have for a long time. I think that's one of the beautiful things about growing up in the Bay Area is like you see how technology has such massive impact right and and then you think about okay well what are the sectors where that actually can have far-reaching you know sort of uh sort of longer term impact and energy is one of those right so back in uh really in 2015 is when I started to spend some more time specifically working kind of in the oil and gas space and this was a function of I got introduced uh by a a very good friend from business school and and now an investor and board member um to a small oil and gas operator that was based in Panka City, Oklahoma that was trying to figure out how they could optimize production. And I had worked in technology before. I'd been on the investing side previously. I'd kind of gone through during business school. I was always trying to figure out new businesses to start. Sure. and and I started to see where there were opportunities to use more of these kind of advanced like hardware solutions that were coming out that were most notably kind of in phones, right? So you had different types of you had magnometers, you had different accelerometers, you had all sorts of stuff that's now sitting it's like supercomputer effectively in your hand. Can you start to use some of that technology plus advances in data science plus the reductions in costs in compute power and storage? What could you do with that? And through this introduction, we ended up realizing that we had a technology that was able to understand movement in a very precise manner. And and what we did was we took Android devices. Mhm. We put them in basically plastic boxes and we zip dive them to pump jacks out [snorts] in, you know, out near Panka City. Yeah. And we started to find interesting movement patterns in that. And what we realized was if you could correlate the different movement pattern to different downhole conditions, you were [clears throat] sort of starting to be able to do what some of those, you know, old school operators do when they walk out there and they put their hands on the pump and they listen and they understand like there's a whole bunch of signal that comes through that. Yeah. We're like, cool. Can we capture some of those details? And we started to we we built from that point. Yeah. And fast forward, what we realized was that was helpful. like the hardware stuff was interesting but uh as I like to say they don't call it you know they don't call it easy wear right yeah so [clears throat] um there's a piece of this which is can we do more and more of this originally pattern matching into so some sort of a pattern leads to some sort of an insight and then from that insight do you take an action right and and Kelvin today is really all about taking insights to actions and and we focus on this concept of we we want to be able our mission is to deliver autonomous operations. Okay. To do that you have to start with applications optimizing assets. And so that's what is an application doing? It's taking a series of inputs from what could be the data streams that are coming off of a particular well and using that to find the patterns which inform some sort of an action speed change or otherwise. Yeah. And what we've realized is that you don't need to that's that concept of I have inputs, I have logic, and I have outputs like that that becomes something that's extensible across a whole bunch of different parts. Yeah. Of an operation. And in many ways, it's kind of what the engineers do on a day-to-day basis, right? Um so we look at it and we say, what does your best engineer do on their best day? Right. Yeah. And typically what they're doing is like they're synthesizing some information Right. They're deriving an insight, right? They're taking an action and they're measuring that impact. Yeah. Okay. That's like that's a day-to-day engineering workflow. Can you do more and more of that actually in software, right? And can you make that what they're doing as kind of a series of manual processes? Can you turn that into something that runs consistently and more automatically? Yeah. And and that's what Kelvin has really evolved into. So we started with pattern matching interesting sort of signals coming out of those pumps to today looking for those signals but turning those into insights and then taking actions and and through that process we are delivering on this concept of moving towards autonomous operations right and and as you know it's kind of from a data science background the more that you're capturing those work streams the more that you're able to measure those impact the better you're able to then predict what's next. Yeah. And you can start to learn and you can start to connect more of these things together not just as individual assets but as part of a broader system. Sure. So can you can kind of bring that in then that how that improvement then helps with buyin because I mean like again or we can maybe just talk about buyin with AI ML just in general and because that's a good talk about you're working with the guys that think you're going to replace them. Yeah. So I mean so if you go back into our history sort of 2015 2016 a lot of our initial approach was hey we have we have an algorithm right we have an approach that that we think is going to be able to optimize this particular asset type and we're very good in terms of getting buyin from senior level folks and then we sit down with folks who [snorts] are like wait a second this does what I do this is a threat and in that situation you get a very natural reaction where folks are going to try to sabotage what you do, right? Makes sense. Um, what we realized though is that rather than coming up with an alternative, you know, and hey, here's a better model, can we instead look to figure out who's who on your team is doing it the best. Yeah. Because if we can capture and codify what your best engineer is doing on their best day, that all of a sudden changes the dynamic. Yeah. Because now they're like, "Oh, wait a second. I a personally get leverage because I'm the best at this and I don't like to do all these manual tasks. So if I can describe it, turn it into an application, it frees me up. So I get some benefit, but the next piece then becomes maybe the other guy who's pretty good thinks he's better, right? So James thinks he's better than John and James is like, "Well, I can build a better algorithm. I can build a better application. Let me show you." And you get this interesting dynamic where versus us coming in and saying, "Hey, we can do it better." We're like, "No, no, you can do it really well." In fact, the constraint is often times the fact that the best engineers don't have enough time, right? Yeah. It's 24 hours in a day. They have 24 hours in a day. So, you know, so the common refrain that we have is uh you look at all the objectives from the CEO. So, I'd listen to a lot of earnings reports and I've been doing, you know, talk about using Genai. I use Genai to go through and like synthesize earnings releases. Yeah. What are the CEOs talking about? And there's sort of four things that we hear about on a pretty consistent basis. Need to increase production. Need to reduce cost. I need to achieve my ESG objectives. Yeah. Right. And I need to maintain safety. Well, that's a lot for one. [laughter] Like those are those are conflicting objectives if you think in certain cases. Yeah. Can be and and how do you balance, right? How do you think about that? And that's where we started to realize that if you can put those types of objectives into an application, you've then freed that person up to be able to work on higher order problems. So you're getting a lot more leverage and you become a force multiplier 100% for the engineering team. And that's really where we have come at this now. So originally it was a much more adversarial sort of for sale and now we're like guys we know you're buried. Yeah. Right. Like you don't have enough time to get your stuff done. what can we what can we do to help you and and because it's now so easy to build an application and that sounds like a big word but really it's it's if this then that in certain cases to start that's AI in general right well I mean necessitive if statements that's right [laughter] so there I mean in terms of sort of demystifying some of the AI piece you know we talk about this a lot with with customers AI is not evil right nor is AI omnisient right AI is going to be a tool, right, for you and and the the thing to do is to understand how to use it appropriately. Okay? And if we can help your existing engineering team get leverage out of the fact that AI is now a useful tool, it's not it's not going to take away their job. It's going to free them up to be able to do more and and to and to again focus on higher order problems. Yeah. Yeah. I mean, I think there's a couple things I want to hit on there. I mean, a like definitely a force multiplier. I mean, talking about Gen AI, you know, say I've got GitHub Copilot running my IDE. Now I can write a script in five or 10 minutes that would maybe take me an hour just to hack together, right? You know, before. So again, frees up more time. So I mean, y'all are trying to play that role, it sounds like, in the engineering space essentially. Yeah. And and our sort of our typical users are kind of are two two of those categories. So we think about production engineers who are the ones who are utilizing the applications that are that have been created. And we have software engineers who are oftentimes building applications. Now that person could be a data scientist or otherwise but typically they have some understanding of how to develop in Python or at least a basic you know sense for sort of how to set up a model and then what we try to do is is create a really easy way to take that model and we talk about like every model really wants to be an actor, right? Yeah. There's a lot more money in it. [laughter] Okay. Fair enough. Yeah. So um so so how do you how do you make that conversion and when you do then it becomes interesting because we know that there's a ton of unutilized models right there's a lot of stuff that's sitting on the shelf so can you unlock those can you empower them can you can you put them into real operating environments in a safe secure and scalable way yeah yeah right and I think another one thing I want to get on there too because it seemed like you're talking you know you take say your best engineer best couple engineers and So then you're able to institutionalize that knowledge which is a you know I think a big thing for me too. You know we're always trying to recycle. All right you've got that in a PowerBI or a Spotfire project. I want to pull that logic back into our data warehouse where it's source controlled. Totally. And then again it's a proverbial Johnny gets hit by a bus thing. Y but I think another thing too is being if it's wrong is wrong in the same direction. I think I think that's under understated like you know like a really bad golfer slices and hooks and you have no idea how to fix it. But if I only hook, I can know what the fix for the hook is and I can straighten it out. But like if I'm all over the place, right, left, you're screwed, you know? Like, so I mean, like at least with this, like there's some directionality. Like if the models are veering off, they're going to veer off together, you know, in the same direction in theory, most in most cases. I think that's that's often times the case, right? Um, but I think this is where again getting into the buyin, right? you know, if you've got if you got [clears throat] folks who are on the production side who are saying, "Hey, I've got a I'm crunched. I don't have enough time." Yeah. And and I could see how these tool these tools are valuable. Yeah. That's great. Like, let's let's try it and let's also do it in such a way that it's not cost prohibitive and it doesn't take, you know, nine months of of meetings to kind of get something up and running. I think one of the other pieces around your buyin question is it's just a lot easier today to get this stuff implemented. Yeah. And and that's a huge shift both I think psychologically I think folks are more open to you know different types of technologies being useful and two the technology itself is just way way more powerful and easier to implement and you know you're using your git GitHub co-pilot to build stuff there are going there are already and we've done some of this work already to be able to say like we can build similar tools to help that production engineer like I got to go synthesize you know I got all these documents I got to get Yeah. Like, and I can I can do three well reviews with my geologist, with my um reservoir engineer in a day. Yeah. Like, can I can I up that to 10? Like those types of things are where we're starting to see again very targeted applications of different types of AI. And and AI again, it's this catchall, right? And I I like to be a lot more precise about what we're talking about. So as it relates to Gen AI, I think there are some really interesting tools around document synthesis. Yeah. But it's great. It's great for batch data. Can't use streaming data right through those systems right now. But can I go through and have it look, you know, be able to understand, okay, analyze these different documents, give me some signal from that. That saves me time. Great. Yeah. Yeah. Pump cards and Yeah. Yeah. I think the pump card stuff is going to be interesting. I'm even seeing I just look at like go through the existing documents and tell me about am I in compliance you know across all of these different wells like do I have anything that I need do I do I need to you know increase my you know pumping at this particular well like those types of things are even simpler examples to start and then what we'll find are more and more opportunities to use these tools the the truth though the biggest challenge that we've seen is many of the legal departments are still unsure as to how these AI tools are going to get used. And so we've seen in certain cases folks just shut stuff down. Um and and that may hold, you know, I think we're as as folks are more skeptical around how this works. Yeah. There may not be like rapid rapid adoption, but the the major cloud players are all pushing Oh, yeah. to to get better access. Well, and at least with some of that too, I mean, I know obviously Microsoft is investing huge, you know, hugely in open AI, but I mean like they've already got stuff where you can create your own vector database within the Azure platform and it's governed and secure. Totally. Um, I mean, data bricks, I mean, you name it. I mean, they're all they're always saying they are anyways, too. Yeah. Um, but I think it's also some of that whole crossing the chasm thing, right? I mean, like some people are just going to wait and see, and they're going to wait until other people have proven it out and other people will reap the benefits earlier or maybe fall in the sword. Who knows? Yeah. what most of my my edge stint at Hivesell made me truly realize that oil and gas is not the only uh the only industry that is a fast follower. It's pretty much every every major industry has that's crossing the chasm is such a good book because of that because it doesn't matter what industry it is. Technology adoption is the same. Y everyone doesn't want to for the most part put their necks out and be the first one to risk it and then fail and fall on their face and then all that stuff. But once someone has something and you can come to a customer and say we're working for so and so, it's amazing how much quicker those next meetings come totally. Just because you're now working for whoever [clears throat] and it's uh yeah, it's a it's a weird dichotomy. I I also think a lot of a lot of the initial, you know, 5 10 years ago even um a lot of the tech companies that were coming into the energy space were coming at it from a purely technology side. Yes. And whether they knew it or not, they weren't like to your point, just how you switched how you framed the conversation with the clients completely changes how they approach you. Absolutely. And their reception of it and their perception of it. And so being able to, you know, engineers like to think they're the smartest people in the room in most cases because they are [laughter] all of them. And uh and so by instead of coming with them with hey we've got this custom model that you know I can't explain to you how it works and you'll never understand how it works but I promise you it'll work to you guys have the models or you guys can build the models or you guys can guide us on the models and we'll help you implement scale train deploy manage y makes a complete difference right like because anytime there's there's uh involve involvement from the end user, the buyin is so much easier. Like totally, I learned that in industrial engineering, but uh it's uh working on the factory floor. It's like, hey, we're trying to make improvements to shave a minute off of the production time of this thing. Yeah. Well, guess what? The first people you should go talk to are the people that are doing the production of that thing. I guarantee you if they've probably tried a bunch of things that you're going to come up with out of the gate, right? Like it's Yeah. I think there has been historically like a lot of hubris as folks have come into this market with the hey we know better and listen I I guarantee you there were people in this market who would recuse us of that you know historically and and I think what we've really tried to do is is is switch our perspective and our and the way that we approach with a lot more empathy right which is what you have real issues that you're trying to to solve. Yeah. Can we understand those? Yeah. Right. And once we understand them, maybe our technology is a useful tool, maybe it's not, but at least you can articulate what those issues are to us. And if we're not the right, you know, if we're not the right partner for you, we can probably direct you to somebody who's a better fit. Sure. That's a that's a different way of approaching this market than I think certainly a lot I mean there's a lot of folks who have tried to bring advanced software, AI, fill in the blank into the space and have just failed. Well, and you know, then you get the the push from investors and stuff, right? and they see the giant market cap for energy and they're like, "Oh yeah, that we want to be in that space." Until they get in that space and they realize that, you know, it's going to be two to three years before you do any work for a super major just because that's how they work. If you can find the right person, right? Well, that's the thing. You're going to have to find the right person. Like, oh well, if we can just land Shell or Chevron or one of these giant, it's like, you're going to spend three years doing it. And it's that's just the life cycle it takes. Instead of going and targeting some of the smaller independents or the smaller PEC companies who their objective is to reduce like they are much more incentivized directly to implement things that are going to drop costs, improve production, improve efficiency because their whole plan is an exit, right? The bottom line. Yeah. Yeah. The bottom line is what so much more important because everyone wins on. But you know what's interesting that's even evolved in the last few years, right? Because when I first started, lots of, you know, PE conversations and folks are like, well, we all we're going to do is punch a few holes in the ground and flip it to the next. Yeah, there's that, too. Yeah. That's changed a lot. And so that I think what we're they're not even going to waste time and effort implementing this new technology because we might not be here in six months. We're going to be out and [clears throat] and we're not going to we're just going to we're going to go with whatever the management team wants because we don't want to be too hands-on, right? And then we've seen that shift, right? And now all of a sudden it's in certain cases it's taken a little bit more. I think some of the private equity folks are taking on what we've seen in the you know in the Bay Area around the venture model which is a lot more services right that they're bringing. So hey we can help you with talent. We can help you with marketing. Yeah. Some of the PE folks you know that we're talking to have said yeah we can help you with how to optimize this. Now you're going to be holding this for a lot longer period of time. What can we do to actually share knowledge across different portfolio companies? And by the way, it's also good for the certain PE shops that might actually have to report on ESG stuff and some other things. It's like is this information available? Are you able to impact decisions? Those kinds of things. Yeah. Yeah. Can you talk about kind of I mean so you mentioned some of the you know the hubris and stuff and I think people can come in and then they think they can solve every problem and it seems like you guys have kind of narrowed to production but I know you've dabbled in other sides. I mean, so I mean just like how has how's that evolved for Kelvin and how like I mean where have you guys kind of found the sweet spot? Yeah. So for us some lessons learned too because that's I think I mean as someone that has been in product for the last few years in the energy tech scene. Uh I think that's that's a very tricky thing, right? Because it's hey the there is this kind of tech boom in the energy space because there is so much opportunity like you've talked about like there's so much but then there's also you know it's like well which one do I go with? And then like you've got to there's that balance of focusing and and going all in the same direction but also as we grow we need to continue to add new things if we want to keep growing kind of mentality. And so The key is to find the pain. Yeah. Right. And and for us the pain initially was production, right? Where folks were saying don't have enough engineers, don't have enough time, want to be able to increase these things. And so that that's where we started and we found we were fortunate, right? We had some early wins and we found sponsorship coming out of there was a group of folks who, you know, at BP who who got like excited about what we were doing and were in a position where they really wanted to affect change. Okay. And there's a lot of serendipity in that. Sure. And but it was about being able to really support what their initiatives were with a technology that could actually have real impact because I think this is where when you're selling to engineers, you have to it's like identify the pain, but then you have to deliver a solution, right? And and if you do that, then you curry favor and you get more and more opportunity. And we were able to do that, you know, with with that team over time. And what we found though was our business as I mentioned had hardware and services and software and other things which are just they don't scale as well and there are a lot of dependencies. There's also generally speaking they're also very different business models like field service hardware and software are all standalone businesses as they exactly and field I mean don't even get into like the insurance and the things you have to have even if you're on the surface but technically things [clears throat] go downhole. I mean it's like it's insane. I'm not going to get into that rant today. [laughter] So, so one of the things I I met with a very large controls company that told me this was in 2019. They're like, "We are we're getting out of the hardware business." I said, "That's odd." Like, "You're a huge hardware provider." And they're like, "Yeah, we're not big enough scale. We're not like Foxcon, Apple," and I said, "If they're too small, what am I doing [laughter] in this space?" I mean, this is ridiculous. So then it it really forced us to rethink what we had in terms of you know a real understanding from a software point of view and where we were strongest and and at that point in time there was also an awareness that we needed to continue to upgrade our team on the control side and also on really building enterprisegrade software given the types of companies that we were going after. And and so that meant we had to we brought on some new people and we changed up the leadership team and we upgraded a lot of these key positions in areas that we knew were going to be important. And and then what happened was we we were fortunate because we had demonstrated some success already about being able to get software to work in remote challenging environments. Okay. And we had a large like like nobody that we had marketed to a large company come to us and say we have a problem. We are in a position where we've developed a lot of really interesting software and applications and we know it works like we've we've we've done it but we've been cranking on how do we actually get this stuff operationalized and into the field for two years and it's not it's not working. And we said, "We got you." And and at that point, we went from being purely in production to starting to move into really into the completion side of the house and starting to realize that their problems were not so unique relative to what we'd seen on the production. You know, they were more complicated because you're talking about, you know, moving crews around and lots of dynamics and complexity there. But you could still distill it back down to this point about, oh, you have an engineer who you have some expert Frack engineer who's decided they have an approach for this particular part of Frack that they're best in class at. Great. Let's put that into an application and like let's help them be able to then get that into the field and run and be successful. And that has led to both continued opportunities there as well as on the production side. the acute pain of I recognize I I know I have a solution for this but I can't I can't deploy it successfully and scale it that has been pretty critical for us and finding those opportunities and then once you do that then it builds right and you end up with folks who are saying hey I I see you did it in one part of like we have you know with we've talked about Santos that's you know sort of what we've done there is public like we're doing a lot of work with them but that started really with some pretty simple initial production optimization application and now today we're you know we're moving really across the enterprise. Yeah. And and that's been because we have been able to demonstrate we understand what the engineering problem is. We can help them put that into an application and then we can support that at scale depending on wherever they want to put it. Yeah. So, I've got a an interesting question because we've got a lot of similar experience as far as the data field side of things. What if what advice would you give to somebody because the oil field I think one of the things that a lot of people underestimate coming in from outside of the oil field is that the oil field is unique around uh the fact that the operators own all of the data, but they generate next to none of it, right? And so like you know if you're Google it's very easy for you to control the input and the requirements and y all of the data that you're then turning around and analyzing and throwing at models. The problem with especially on the drilling and completion side is the fact that all of the data that's generated over there is generated by dozens of different service companies and different formats and different standards etc. at different timestamps and intervals and all that fun stuff. Yeah. Um, and I my hypothesis was the reason y'all had it's not just y'all. There's been a number of companies that have started in software got into the hardware space in my opinion as a means to an end because there was no good way to get the data into the software. Y and then realized how difficult it's not even necessarily hardware, but it's the field and the hardware component of all of that is and how obnoxious there are obnoxious things like insurance are and you know all the safety requirements to have guys out in the field and all that stuff. Um, can you just like unpack that a little bit because I think there's a lot of people that, you know, are have an idea for a piece of software or whatever, but then it's like, well, if I only had this data or if I only had a way to pull data from this thing or know this specific thing and you you guys have a unique kind of you specifically have a very unique experience around that. So, kind kind of talk about that. Sure. Yeah. So I think as you've articulated your software is only going to be as good as the data that goes into it in terms of its ability to take appropriate action. Okay. So for us what we realized was initially yeah we're going to go gather that data sort of raw from that particular pump and use it to define different types of insights and actions. Fast forward. What we also then recognize though is there are multiple ways to gather that data today that may not have existed three years ago, four years ago, etc. So what we will find is there probably does if you're talking about working in a remote environment, you can't rely on the cloud. Okay. So that's like I want I want to get in on Yeah. So let's talk we can talk more about that. So, so there's there's just way too much interference, wireless problems, you know, you've got lots of things that are happening in the Perian today that will impact any [clears throat] sort of like cloud system having true streaming data that's that's pristine. And no operator will ever sign off on any kind of automated or remote control if it's cloud-based because of the fact that yes, if it goes down 0.01% 01% of the time that's enough that's enough that yeah the risk is so high in the field that that's enough that it's that's right and so that's led to this real need and and kind of where we've spent a lot of our time is this hybrid world between edge and cloud okay and so a typical way that we would approach that is to say okay well what operation are we talking about if we're talking about completions then you're probably going to want to have some additional compute that's sitting in you know in the van right and That's that's pretty straightforward because there's already compute that's out there. Maybe you add another box, but it's going to be a you know a generic Linux box that's hardened or you you partition something that's already there and you use that compute and storage. Great. So now you're getting a much more direct source of information and you can sort of measure the purity of that to determine how good it is and you can you know do your analysis to make sure that it's good enough to use for something else. you're doing that at the source. And that's a big thing is part of the reason we've designed our product the way we have is we want to be as close to the source, not just to read the information because we're a controls company. Like we need to write back, right? So so having the ability to get access to hardware that's going to give you the information you need can be yours. You can partner with somebody else, but that's that's critical. Sure. Once you have that, what we find is most of the time you're going to be making changes as close as you need to at, you know, to that system. So that could be maybe you are making control changes at the edge. If you're doing that near real time or on a streaming basis, that totally makes sense. If you're making a change once a day, maybe you can run that out of the cloud. Okay. If you don't, so so that time sensitivity depends on the nature of the the control decision that you're making. Yeah. Okay. So, it gives you flexibility then to figure out because the closer you get to the edge, the more expensive your compute and storage is, right? Yeah. Versus you're you're in the cloud, comput storage is cheap. You're at the edge, comput and storage [clears throat] is expensive. And so there's a calculation as to where you want your application to run, right? So if you can build a really sophisticated interesting model but distill it down to a rule you know E= MC² you can put E equals MC² an application run at the edge. Yeah real easy right if you need to run train and do all these you got to do that at the cloud. So it's this recognition that you want to be flexible and not try to force all of your models to the edge. You don't need to do that. Yeah. You need to you need to decide what are the decisions you're making and put those as close to that to that asset as you need to. Yeah, I think that's a big uh question I would say around edge with a lot of people is it's like well what do I do here? like why do I do it here versus there and all that stuff. And to your point like people don't think about like the urgency of a decision essentially and it's like if you're trying to make a recommendation or something on a operation that's happening with people on it in real time that has to be done as close to the data as it possibly can at least in the oil field. Yep. Um but you know like you said if you're opening and closing a valve once a night or checking if the condition of something y probably not that urgent, right? Yeah. It comes down to that that definition of real time, right? Like some of these real time things are literally micros secondsonds and some of them some people mean in 10 minutes or a day is real time for some people depending on what they're doing. You mean I was going to say especially from like the SCADA side or production and SCADA to operation or uh drilling and completions, right? Real time is very different like 15 minutes is real time in skater world but real time in completions is 1 second and then real time on hardware is nanoseconds milliseconds right that's right so it's uh it's an interesting thing but yeah I think you know the the two like big things for me going into the edge space and and all that from the energy side was you know that ability to have basically uh what is it called I'll remember the name in a minute But controls in a disconnected environment. Yes. And which essentially means that the next step of that is automated controls in a disconnected environment. So that's the the future automating some of these things truly automating them in the field and letting them run. And that's that's really what we've been able to do. But then the next thing is federated learning with some of that stuff where that get like that people don't a lot of people struggle with what that is. But like when you look like chat GBT is the best use case that the average person understands about federated learning right like millions of people downloaded it really quick lots of people are using it and they have continuously improved that model month over month over week because it's continuously being trained on individual edge case models essentially right and so that's uh the federated learning stuff is just I think that's going to be us coming out of the trough of despair for a IML is like people getting these federated models really down, getting them into production and just seeing how quickly they're going to be able to accelerate the learnings of them. Well, there's some other there's an interesting sort of extension to that as it relates to the kind of the oil and gas space which is historically we've thought about the different parts of the production process in in pretty siloed like you know sort of entities right you know I may have SCADA that runs across my whole system but I'm using one piece of software to optimize a certain lift type I'm using another you know I'm using some other software to kind of manage potentially compressors I've got other stuff that's managing distribution and and these are not all tied together in a in a thoughtful way. And one of the things that we we have been able to show is if you can look at these different assets as part of a system and not just thinking about them as a series of like local optimizations, but rather like it's our global optima, there's tremendous gains you can get out of that. and and when it works, it's it's that sort of thing where people can't quite believe it, but it we've seen this in so many other spaces, right? So, for me, I always look at financial services as kind of a a parallel where it's like you think about programmatic trading. Yeah. No, they've had Right. That wasn't a thing and and it's high stakes because that's one thing we always kind of knock like these people come with their AI and they don't understand the risks. But like automated trading, I mean like you make the wrong thing, you cost someone millions, billions of dollars. I mean like or crash the market to shut [laughter] it down, right? So, so I look at that I'm like, "Okay, well that's that's system level control at scale." Yeah. Right. And again, faster than anything we're talking about here, but we know that those models exist and we know that they work because we've seen people make billions of dollars using that. And it's and over time as you start to bring some of those concepts, those network concepts into these types of systems, you're going to be able to unlock some really interesting things. And so you talk about federated learning. I think about that also as there's some interesting causal inferencing modeling and some stuff there where you're getting into a world where yes catch PT is awesome and there's some really cool tools around that. I'm not sure that's going to have the same transformational effect on this sector as you will see some of these other fields of machine learning and emerging you know categories of AI because the beauty of control is you know every time you make a decision what the impact is right and you can learn every single time what that is and I look at that I say we're going to we're going to be able to open up you're saying when it's a system like that you know if if you know even if my pipe diameter goes from here to here that like there's this different pressure temperature you Pvt conditions that are affected by it, you know. I mean, you're you're talking about production that obviously impacts your marketing and mid-stream side. I mean, um, that's a I mean, it's funny because it's it's not just, you know, production or that side of the space either, right? Like completion. I mean, frack interference is like the perfect example of this that I first noticed, right? It's like I don't know how this probably damn near 10 years ago. I went to a conversation that uh it was some lunch and Dave Kramer, shout out to Dave Kramer uh Ed Kico did, but he was talking about the they were doing new wells and like uh it was yeah new wells in a semi- depleted area of the Barnett and they were having so much trouble fracking them and no one could figure out why and they were just looking at the individual fracks or wells and So, I don't I don't remember how we ended up started looking at them, but Dave took a step back, looked at all the wells on the map, and we're like, well, these two wells over here have been producing for the last 2 years, and they're reservoir pressures are super low compared to the normal reservoir pressure of the Barnett, whatever that may be. And so, he started just thinking about it physically and holistically. He's like, well, they're just pressure sinks. Like, they're just sucking literally sucking the fluid away from what they're trying to stimulate. And so, guess what? They shut in those two wells. They I think they ended up loading them up and stuff, building the pressure back up, and then they had no problems fracking the well in between them because it no longer had this normal physical phenomenon known as pressure. And so like the fact that you know even just that's a completions engineer on a you know not even a pad but just a small part of the field. Yeah. But like the fact that the oil industry as a whole, we don't no one because of how it's segmented, no one thinks about how the drilling schedule impacts the frack schedule, impacts the the flowback schedule, impacts the production schedule, impacts the marketing, impacts the sales, impacts the refiner. Like it literally impacts the entire thing and anything along that way has impacts both upstream and downstream of it. That's right. And there's but you've got contracts, you got to meet. So it's like it's this giant dynamic multivariate system, but most people don't think about it that way. They don't model it that way. And so it's really cool that y'all are pulling. We believe you have to you have to put this in place starting small, right? You start with production and then you get into midstream and then you get into other parts of it and you begin to sort of build this up. Um I like to think about this as much more of a biological system, right? is a series of cells that you're just saying like we're going to connect these together and like the production cell is going to talk to this cell over here on the distribution side and they're going to connect that and like that's what we what again very high level but that's how we conceptualize it. You touched on something else though that I think kind of tying back to AI that I also wanted to highlight which is there's also an awareness that AI and sort of data science alone is not going to be what solves all the problems in this world. Like we have to think about this in a much more hybrid model approach, right? You need to be able to think about how do I take a physics-based model and how do I use it potentially and augment it with something else like th those hybrid approaches are possible today. And and that's the other piece that I think we're going to start to see maybe some unlock, which is okay, I've taken this this model that I had that was pretty good and then but I've been able to modify it to some degree to turn it into something that is a little bit more data enhanced and see what those results are. Yeah. True. And and that's kind of where I think again a lot of this is hybrid. It's not going to be black and white binary super clear. It's going to be more of these blends that you're going to bring up. Yeah. that you know people again people think that AI is going to replace their jobs and stuff and it's like for the short and near term AI is just going to enhance right it's going to streamline trim time like all of that stuff like it and you know Freriedman or uh yeah Freriedman yeah he was on the Rogan podcast talking about you know he's like everyone thought we were going to take blue collar jobs with AI out of the gate right and it's like it's actually just enhancing all of these other jobs that are more creative or whatever or and and it's not taking them, it's just making them easier, faster, more streamlined. It's a tool. And maybe this is too frank, right? But it's like if AI replaces your job in the next year or two, you're not very good at your job. That's I mean like learn the AI and become an expert in it and even prompting right like it's funny. I have a friend, this is outside of this space, but I have a friend who uh he's a lawyer uh sorry he's a is a law professor at University of Colorado and his name is Harry Seren. He's done a lot of work in AI applications to the law and all over LinkedIn. Lots of activity and we had breakfast a couple months ago and he said, "I just have a my cousin is graduating from college right now." And I'm like, "Oh, that's interesting." He goes, "Do you know what [clears throat] I gave him?" Like, "No." He's like, "My graduation gift to him was an OpenAI subscription." [laughter] That's awesome. because he said that if you understand this and you actually learn these tools, you're going to be so much better regardless of what job you go after next. But just just figure out how to get some mastery of this and see where you can go to. And I thought that was insightful, right? Because it's a tool, right? And it's not scary if you actually start to use it. It's actually really empowering. You can understand the bounds of it, too. Like what it's not good at, what it is good at. And you know that's a like the one of the examples I used in my pitch deck was uh was for breast cancer screening right doctors alone were like 80% accurate the ML it was a computer vision model the computer vision model uh of the I think their MRIs was uh like 75% whatever but combined it was over 90%. Y and that's literally millions of lives of people who were saved every year just from that. It's like the hybrid approach is absolutely I think I think people haven't been told or it like people don't talk about that a lot in public or zero sum it's all it's this all or nothing but that's absolutely not how anything works how this works right uh and so I think that's where there's so much potential is if we can stay listen I'm I'm an optimist and I believe that there's these are really powerful technologies that can be used for good right and now it's a question of how do you make that tool appropriate for the job at hand and then once you've done that it will open up new opportunities and new cool things to do and you know on down the line. I mean like I was just think like like the brain computer is what we're trying to go after like so you're always going to be trying to get there. I mean so at some point the brain is still going to be better or have more capacity you know now different models can maybe be better at more narrow scope things and I think that's what we're talking about here like when you combine like just the overall general computing power of our brains but then focus it with something more narrow focused you know AI then like that that is really really good at that one thing or a couple things and you we're better at learning like higher level things than we are like because we're not able to process like millions like like doing like ridiculous math, you know, and even adding two numbers that are huge. Like we can't do that. Yeah. Off top of our heads very quickly, but the calculator can. But understanding that, you know, thousands plus thousands equals thousands versus, you know, those types of things are are uh and I think the human brain is incredible at pattern matching. Yeah. Yeah. Particularly AC, I mean, where where does AI struggle? It struggles like if I go from text out of a PDF to, you know, text out of a different file type, it may struggle, right? Maybe today it's better, but human brain's going to be able to figure out patterns across those. That's like we were talking about that B uh model that Bobby had had written and you know we were trying to do it numerically and I told him I was like it's annoying because with frack like I can sit and look at any treatment plot and tell you exactly what's happening during the treatment. But programming that into a computer is so hard, right? How am I going to write that function to actually like so many edge cases? Yeah, you know, which I mean that's the that's the problem with AI is or ML in general is you've got to program all the you've got to get those edge cases people you can you have good data sets but you don't just want good data but then it's only as good as the data if you've given it like you want good data that also has bad things that happened in it so that you can find yeah and train yeah [snorts] but I think that's where you're seeing like really interesting work and there's going to be better and better training approaches and things going forward but you're not going to take that human understanding like when we talk about Kelvin we talk about I want to I want human expertise plus artificial intelligence to yield better you know impact right if you do that you're doing some interesting work for sure and and that's applicable across all sorts of are right as well as more broad domains so what's uh give us some some real world examples that you can um as far as you know some of the impacts you guys have had or just the the findings that you guys have had on on some with some of your Yeah. So recently what we've been able to do is I think this is a classic example of like some of the so with Santos there was a a particular engineer who was an expert in progressive cavity pumps. Okay. And his approach was, I want to be able to take some information that comes out of the reservoir and I want to be able to also use information that's coming off of the pumps themselves and and with that in Excel, I can start to actually help make better decisions around how to manage this stuff. And we looked at that and we're like, oh, that's a perfect, you know, potential application. And there's there's some real data science in some of that to do pattern matching. And there's some real like sort of human understanding of that process. So if we can capture and codify that, we can turn that into something that can be quite powerful. And so that was very appealing to some folks in Australia who have, you know, a lot of progressive cavity pumps. And so what we did is we just started with a typical kind of trial and began to run that and be to sort of see what the results were. And very quickly it started to show production gains that were meaningful. It started to show energy reduction because we were running the pumps more efficiently, reduced number of works and all of the goodness. Yeah. And what that has led to is sort of two things. One, expansion of that particular application. So that's now running, you know, across over a thousand wells. So that's been good growth. The other thing though is it's really inspired the you know the folks at Santos to start where other what other things can we apply this to this approach right which is yeah we have a great engineer who has this awesome idea and so we're working on actually now six or seven different projects okay where that concept can be applied and it can be done in the beginning at a small scale right and if it works great but I think it doesn't always have to work it doesn't always work so then the question is how do you learn from at and try a different approach or do you go after something entirely different? Sure. So that's on the that's an example on the production side. You know on as it relates to sort of some of the work in the OFS world what we are finding is that you break down you can break down sort of the completions process similarly and there are great engineers who really understand how to optimize certain parts of the of the frack process. And in that case, what we've been able to show with them is like this can have a material impact both in terms of are we are we fracking to plan, right? Are we getting the results that we expect, right? Are we doing it as efficiently as we possibly could and is it being done consistently? So these are all things where historically you can get different crews who have different approaches who are thinking about hey I've got a slightly better way to do it and I'm going to kind of take take my own sort of understanding of this situation and apply it. Okay, cool. But is that really what is that really what the boss wants? Yeah. Right. I don't know. Maybe when you I mean you can see those biases between the morning crew and the night crew. I mean like I mean you know one guy's way more conservative and taking longer to get to breakdown. another one rips it open and is over capitalizing on horsepower. I mean, there's so many different ways to think about it. Well, that's the other part, too, right? Is it's like what are your how are your contracts written? Are they based off of horsepower? Are they based off of, you know, time, number of stages? Like, there's so many things in there that get really really interesting really quickly. So, it's it's I'm excited for the fact that y'all are not just focused on, you know, the process, but also the other things that tie into that, right? Like I mean that's always been the thing with frack is it's like yeah we could frack more or we could frack you know at a lower rate which depending on my pricing model puts me into a cheaper price but does that still achieving the same frack you know characteristics that I want vice versa how do you know how to do that in real time like all those fun things Bobby and I have had lots of conversations I've got a lot of ideas on the logistics piece of the frack and or sand and chemicals that Yep. I mean when I was even when I was a frack engineer over a decade ago, I was a glorified logistics manager. Yeah. And you know, company man soothsayer just make him happy and make sure all the shit's there so that we can frack and we'll f you follow design and listen to the company man, right? But there's all these things on the logistics side where it's like, hey, uh, you know, the sand hauler got tied up for whatever reason. And then now you're short on sand, so you can't start pumping and now you're getting charged MPT. Like again, all those secondary kind of uh knockon effects. Yeah. Effects that that come from all those things. So it's that's really exciting stuff. Yeah. Again, and we feel like there are so many of these examples where you have Yeah. engineers who have great ideas. Yeah. but they don't necessarily have the tools to act on them, right? And and if we can be a provider of tools to help take those insights and turn them into actions, that's a win, right? It's a win for the engineer, it's a win for the operator, and it it really has like huge positive benefits across the industry. So, talk real quick about the kind of productionizing and scalability piece because I know that's that another thing that a lot of people way underestimated. they don't think about until they're there and they're like, "Well, [ __ ] [laughter] we'll get we'll get a Raspberry Pi and yeah, yes, we have some scar tissue around around that topic." Everyone in the Everybody [laughter] who's touched IoT has it. Uh I think you have to start with do you have a do you have a good foundation of information, right? And and that can be if you if you've got redundancy through your data sources, you're going to be probably just fine. So figuring out what that looks like, that's the first piece. The next piece then is do you have a feedback mechanism on whatever that application is that you're running because there needs to be buyin from [clears throat] the engineers because I again I sort of characterize like to some degree builders and users like there are people who are building applications so you may say hey I've got the best you know optimization model ever and I want to use that great you've got a you can create that you can build that and then we can help you deploy it. Yeah, John's got to actually use it, right? Are you gonna use it? And you're only going to use it if you think a it has to show up that it has real positive impact and b it needs to consistently be able to you need to interact with it to such a degree that you trust it. Yeah. Right. Because trust is the thing that people forget is that if you lose trust with any of your operators or otherwise, they'll figure out every which way they can to turn you off. Right. So, so we feel like consistent feedback about how is this performing. So, we talk about when you define an application, you articulate what the inputs are. You have whatever logic you've created and you have outputs. Outputs are speed changes and otherwise, but the other piece is impact metrics. Like what are you measuring, right? Yeah. So what? And and the impact metrics has actually been something that we we've really focused on recently because it takes away this whole conversation of like who gets credit, right? And you're like the [clears throat] the decision was made, we can measure what the result was and it's clear the application was working and it delivered that or somebody turned off the application. Yeah. And it delivered a different result. So now now you've got this really nice log that's auditable. You can go back and you can learn from that. Okay. So that becomes part of the the caring and feeding that needs to happen with these applications. They don't you can't just drop them in there and work forever. So So that means from a resource perspective, if you're the if you're working on that for your company and you're developing it, you're going to have to allocate some amount of time to retrain, right? But you can you can understand when the when those impact metrics go outside of a range that you're comfortable with, [clears throat] that can trigger it. or if John says, "Hey, it's it's not it's not getting me what I need." Yeah. Then that can be another trigger. And all of a sudden, you now have a much more collaborative relationship because I think one of the other things that we've noticed is many of our many of our customers still have silos between the software engineers and the production engineers. Yeah. Right. Let alone a data scientist, let alone like I mean, so you've got all these silos and we're in a day and age where it's super easy to actually collaborate between people. Yeah. But you don't necessarily speak the same language, right? You speak Python, you know, who knows what you speak, but like some combination of Excel and something else. So, if you can to help bring those people together, share a view on what you're trying to achieve, be able to give feedback around how this application is is actually performing, you're going to come up with new ideas. You're going to be able to give him like, hey, every time I see this, there's a problem. Can you can you sort of write the code to do something about that? Yeah. So now you're getting like real value across both production and software engineering for sure. And that's how we sort of see you have to now you don't always have to be the developer. You can use a third party if you want to otherwise but at the end of the day you have to have buy in that this works. Yeah. Because if you don't you'll figure out a way to sabotage it. Yep. Oh that's very true. Yeah. And I think talking about the scaling part is like I've been on projects where like you have some really talented data scientists machine learning engineers. They don't know MLOps like exactly deploying those is is not the same skill set. Totally like so data science, data engineering, actual application development like these are all different areas of expertise and a link what we've been trying to do is to sort of simplify that as much as possible. So our pipeline can be you've written something in Python again you can define you know you can define inputs and outputs and impact metrics. You probably can. Yeah, great. We'll wrap that and push it out to an application. You get to evaluate it. Go. You don't have to. It's It's hard to push a Jupyter notebook into production. I can I I can show you. [laughter] I've got a We have a way now actually. Um so, but I think those are the types of innovations that are also trying to just simplify what is historically been a complex process. And the easier you make it, the more leverage you get. Yeah. No, I think a lot of people think, oh, well, you're in software, so like you can do every like Well, it's like people think I'm in it even, you know, hey, come fix my printer, you know. [laughter] Exactly. I still fix my father-in-law's printer. Yeah. So, I know Bobby and I do the same with all of our relatives. No, but that's uh people I think a lot of it is just we're not the energy industry in general is not nearly as familiar or experienced at building software especially at enterprise production scale. And so it's one of those things that you don't know what you don't know and they're like, "Oh, we can try this out and do it and then they get into it and they're like, "Oh my like there's a guy that has to do the data engineering and then a data scientist and then you've got to have someone that does the architecture and the servers and the management and then the production and management of all of that stuff." Like there's so much more to it. Oh yeah. Then you need a guey that is pseudo useful and intelligently done and you know intuitive in some kind of way and like and that's another person that you need. That's ask me how we know that John. [laughter] But what's interesting about that too is what we're finding is there are still operators that want to build it all. Yeah. And they've got they've got a an ambitious, [clears throat] you know, IT department and they understand how to use Docker and they've got ideas around, you know, open source this and that. And the challenge is like in those cases, you walk in and you're like, "Yeah, you c you could do this. You could do what we're doing over two years." And then you got [clears throat] to maintain it. And then you've got like And is that the highest and best use of your time? Yeah. Right. That's that's the thing I keep coming back to. I'm like you of course you could because if you tell an engineer they can't do something, they're going to figure out figure out how to do it just to spite you. My fastest way to motivation is to tell me I can't do something. Right. [laughter] So you're like you you 100% can do it. That's not the question. The question is should you do it? Like what are you best at? And that's where we come back to comparative advantage. Like we're really good at understanding how to deliver control software. We know that applications can optimize assets if you use our system. Yeah. So, so you don't have to worry about that. We got you. Go figure out how you build more applications because, you know, if you do that, you're going to get more optimized assets. That's great, right? That's a win. So, again, it's trying to figure out exactly what we do well and we can complement what others do well and and for others to recognize that yes, you could do everything, but should you? Yeah. No, I mean I think even within the industry as a whole a lot of people struggle with that, right? like should we just be software or should we do software and hardware or should we do you know like should we just do data or should we do data and soft like right these are firsthand experiences Bobby and I have had and it's one of those things that it's it's getting a lot better because the industry is finally opening up to you know JVS partnerships and like actually working for vendors and operators and more than one player is coming to the table to say hey it makes a lot more sense for me to do the data and you to do the software because I've done data for the last 20 years and you've done software for the last 20 years instead of us trying to compete against each other for this same new solution and be poor poor crappier versions. I did a full stack web developer boot camp. I would not pay me to build a website, [laughter] right? Like I mean but I can speak your language and we can have good intelligent conversations or relatively intelligent which is good like you need some of that because you also need you need IT teams that can can call BS on on folks who are coming in and selling in like the moon and the stars. Our approach now is like there's a lot of really interesting smaller companies that are doing super interesting stuff. Great. We can collaborate with them, right? We like I was with Jose Silva from Zaka earlier, right? And you know, they should be a partner of ours. They are a partner of ours. We worked with them back in the day. Yeah. We're now in a position where they can make recommendations. We can take those recommendations and act on them. Yeah. It's funny. I was I was thinking about them when you were talking about because I mean they're a little higher. they're abstracted away say from the pump side for the most part but like you guys but they really play well totally so but even then you see like you can build a model in Azure studio you can do something in seek you can do these other things and and those become again the logic that can drive an application and but we don't have to we don't we're not going to do everything we just want to do our stuff exceptionally well yeah and that's I I think that's to all the CEOs and entrepreneurs and folks out there That's such a like it's so simple, but it's also very hard when you're a company and you especially in a space where there is so much opportunity because then they start they bring you oh but whoa if you can do this with that then can you do it with this and you need to listen to that and document that and keep that there. But being able to focus and specialize, I feel like there have been way less companies that have failed because of that than have failed because they got way too outside of their specialty and spent a bunch of time and money on stuff that they're not very good at. Y and fell apart. And listen, we're guilty of that. I've chased a whole bunch of shiny things, right? I mean, it happens. It's human nature. Yeah. And and and again, that's where I have to sort of keep coming back to say, okay, what do we what are we really the best at and how do we continue to sort of invest in that and be and be focused on it, but it's hard. It's not it's not it's not something that's easy for me to say no to opportunity, but it's what's required if you're going to actually be successful in the long term. I'm a guy, so I love spinning new stuff out. Like Steve Jobs, something like focus isn't like saying one. It's like saying no to a thousand things that like Yep. Um but you want to jump into a little that hour. I know it was quick. I was not uh ticking earlier. I was trying to find my list of questions that I can no longer find. So um we'll we'll do a mixture of them because I remember a few of them. But um Mac, Linux, or Windows? Linux. You want to throw in just the open-ended ones because I probably can't remember all of mine. [laughter] I was going to ask him a real tough one. Houston or San Francisco? Houston. Oh, spicy. Coming in hot. I like it. Um, GCP, AWS, or Azure? AWS. Who's your favorite social media follow?

Any social media. Yeah, it could be anything. Any top outside the domain, whatever, LinkedIn, Twitter, whatever. Peter Tia. Okay. favorite top three books or just three good let's just go with three good books because top top is hard but Sedartha followed by um what did I just finish the

that's such a hard question for me right now yeah I'm reading like four different books Um,

you have any good tech focused ones that I like 0ero to one? Yeah. Yeah. And um, what about business? You've got an NBA too, right? I do. Investment side. My uh, my college uh, finance professor is a guy named Gary Smith who wrote the investments book that I used that I still refer to. It's just real basic, but it's awesome. And then now he's uh he writes all about the limitations of AI. He's also a statistics guy. Uh and he's hilarious because he just tears stuff down. And and I really enjoy that because it's just it's good perspective. Yeah. On what works, what doesn't work, what are the limitations. And I think that comes back to if you do if you understand statistics, you know that like these things are not all knowing. They're not sentient. [laughter] So, so, so great. Like, think about it that way and then understand how to apply it. Yeah. Uh, Xbox, PlayStation, or Nintendo? None of the above. PC? None of the above. None of the above. No gaming. I'm right there with you. The The only things I ever played even before were sports games. And then, but um it's hard to hard uh hard not to enjoy NBA jam. I've told my kids uh this this totally dates me, but um I was really good at a at a Atari game called River Raid, which was and I had like a top 10 score. Is that the one where you like you're going down a river and the stuff is just moving up the screen at you? I had like a top 10 score on took a Polaroid of it and I sent it in and I'm like I'm done. [laughter] just mic drop. Literally, I just said I said I said I could spend years and years of the rest of my life doing this kind of stuff and that's not that's not the highest and best use of my time. That's I like that you got a top 10. I'm like, all right, that's it. Place that flag. We're good. Onwards. Um favorite open source library

pass. Okay, now you know why we went to multiple choice questions. [laughter] It's hard putting people on the spot. Uh we'll do the last one. What's uh what's some a piece of advice you'd give to either young startup or someone in the in the industry or outside the industry trying to get into the energy tech space. [snorts] It will you have to be so resilient. Yeah, that's a very good piece of advice. And and more resilient than you think you can be, but you can be. Yeah. Right. It's a roller coaster. So, you just and just keep going. Yep. And if you do, you can accomplish great things. And if you don't, like, yeah, you'll fail. But just the resilience is the thing that like continues to show up for me. Yeah. Like, there are so many really brilliant ideas out there that don't get executed on because people just don't have the wherewithal to keep going. Yeah. And if you keep going, you might actually get there. Yeah. Yeah, I mean I'm self-titled startup massochist and uh it's one of those things that ev damn near every startup I've ever worked for has gotten to a point where it's like we're about to run out of money or we're running out of money or we need to find another in whatever it is. Yep. And it's amazing at how just like it seems like sheer luck the successful the ones that make it something happens through pure determination and resiliency and grit that you know they went to a happy hour and they randomly met a guy the week that you know they needed money like it's and it's just all about that constant I'm if I stop that's the only way we're failing. That's right. Like, and it's it's so true. Yeah. That look is when preparation meets opportunity. Absolutely. Like, you just keep doing it. Keep, you know, and it's so hard. Yeah. And I think that's piece is like people don't People never realize how hard it is until they're into it. Sure. But then it becomes so rewarding. Yeah. Just because you're like, "Hey, still going." Yeah. Um it's not as rewarding. I probably shines a lot of perspective though, too. It's just like I mean like you realize you start to really realize what's important. Like I mean that you you have that right support group around you and like that's really what's important and then this is a just a byproduct. Yeah. There's also I think you just have to get comfortable with randomness. Yeah. And that's hard because everybody wants control. I just I just I'm the best at something and I'm going to get rewarded because I'll just outwork everybody else. And there are occasions, right, where that happens. But that's the exception to the rule. Yeah. Instead, I find that so much of it is just being comfortable with the fact that there are so many things you cannot control. And this is what I I talk to my kids about this all the time. I'm like, focus on inputs, not on outputs. Yeah. Like, what can you control? Put your attention [clears throat] there. Yeah. Great. The the stuff that's going to come out of the back end. How well did you prepare for that test will be an input on how you do, but if you get all the questions that you studied, you're going to do great. And if you get none of the questions you studied, you're probably going to fail. There's randomness in that. be comfortable with it. Don't be upset by it. Right. Don't let the one question that you didn't study throw you completely off. Just keep going. Yeah. Yeah. And I mean, I don't know how many companies have started in the energy space and just were completely at the wrong time. Like too early, started and then you hit a bus. Like there's so much of that, right? Like and uh but yeah, you you've got to be resilient because the energy industry is full of very high highs and very low lows. [laughter] Y it's it's part of it. It requires a strong stomach, but it's it's also incredibly important work. Yeah. You know, that's the biggest part, right? Is like energy goes into everything. Yep. Like to manufacture anything, you need energy. Yeah. And uh that's Yeah. That's why these guys started DW. So not enough people talk about or realize it, think about in their daily lives, right? Yeah. Because take it for granted. Absolutely. until until you got, you know, rolling blackouts in California and like ice storms in Texas. I was going to say I was out of power for three days. Yeah, I sto in Texas. I mean, it was crazy how quick we went almost to like felt like what I would imagine third world like like people waiting at the one restaurant that's open like and I and then that's where I sort of see, you know, as we think about demand, right? There's a billion air conditioners are going to come onto the market in the next 10 years. Yeah. Right. Yeah. And then and not to mention all the new GPUs to power all the AI and Exactly. and all that stuff. So that's no that was uh that was one of the first lessons I learned about energy was like hey guess what energy is not going anywhere. Demand for energy is not going the source of that energy may change throughout time but the energy industry is literally never going away like energy is required for human flourishing. Exactly. That's right. Where can uh where can people find you? Kelvin.ai. Cool. Best place best place to go. Awesome. Peter, thanks. Thanks for coming, man. Thanks so much. Great to be with you guys. Appreciate it.

Open the transcript file

Hosts and Guests

Hosts

  • Bobby Neelon host Feed-stated
  • John Kalfayan host Feed-stated

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. EP 90: Jonathan Henderson from Spotfire

    Nov 16, 20261 hr 3 minAudio

    Jonathan Henderson spent years building oil and gas workflows inside Spotfire at Enertel, and now he leads forward deployed engineering at Spotfire itself. He talks about why engineers making seven figure calls still...

    Credit the people on this episode

  2. EP 89: Ryan Yang from GridPilot AI

    Nov 9, 20261 hr 6 minAudio

    Texas woke up to a blue wave, and Ryan Yang, cofounder and CEO of GridPilot, joins John and Bobby to talk about what it means for data centers and the grid. Ryan breaks down the years long interconnection grind, why...

    Credit the people on this episode

  3. EP 88: Kevin Kong from Everstar

    Oct 26, 20261 hr 14 minAudio

    Building a nuclear reactor in America used to take three years. Now the paperwork alone can eat a decade. Kevin Kong, CEO and Founder of Everstar, went from scaling Bird scooters to using AI to untangle nuclear...

    Credit the people on this episode

  4. EP 87: Mahyar (Matt) Mohajer from Blue Spark Energy Inc.

    Oct 19, 202655 minAudio

    Scale builds up downhole and the usual fix means chemicals, pumps, and a pile of equipment on surface. Mahyar (Matt) Mohajer, CTO at Blue Spark Energy Inc, breaks down a wireline tool that fires electric pulses to...

    Credit the people on this episode

  5. EP 86: Samra Nawaz & Vinay Acharya from WellWorth

    Oct 12, 20261 hr 12 minAudio

    Upstream finance teams have been white knuckling giant Excel models for decades, and the husband and wife team behind WellWorth thinks that era is over. CEO Samra Nawaz and co-founder Vinay Acharya talk with John and...

    Credit the people on this episode