AI at the Tactical Edge: Scaling Secure, Warfighter-Centric Tech in Denied Environments | The Pair Program Ep79

Speaker A: Welcome to the PEAR program from HatchPad, the podcast that gives you a front row seat to candid conversations with tech leaders from the startup world. I'm your host Tim Winkler, the creator of Hatchpad. And I'm your other host, Mike Gruen. Join us each episode as we bring together two guests to dissect topics at the intersection of technology, startups and career growth. Welcome back to the Para program. I'm your host Tim Winkler. Uh, today we're going to be diving into a conversation about the future of AI in defense. Specifically, um, how secure air gapped mission First AI is making its way to the tactical edge. Uh, and joining me today are three, uh, guests who are building and scaling exactly that. First we have Tyler Saltzman, Colton Mulkerson, uh, co founders of EdgeRunner AI company, uh, pioneering domain specific on device AI agents for the war fighter. Um, and rounding out today's lineup we also have TJ Rowe, the Chief Revenue Officer at Second Front Systems Co. Reimagining how commercial tech gets securely deployed in national security environments. So Tyler Colton, tj, thanks for hanging out with us on the podcast.

Speaker B: Thanks for having us man.

Speaker A: Good stuff. All right, so yeah, we're going to talk a little bit about, you know, warfighter centric design, uh, Edge Compute, um, scaling startups in the dual use world. Um, so lucky to have all three of your perspectives in the room for this one. Before we dive into the main discussion, we like to kick things off with a little warm up that we call Pair Me up, uh, our version of an icebreaker. We go around the room, share a pairing of two things that just go together. Uh, so I'm going to kick it off, um, I'm going to go with shooting ranges and V8 engines. So I recently went to a local range, uh, in my area, a spot that's called Excal. Uh, and just pulling into the parking lot, literally every single vehicle, including mine, was either a truck or a Jeep, most of which had some sort of like a lift or souped, uh, up in some fashion. Um, but all of which had, you know, certainly some level of heavy horsepower engine. So uh, I'm going to go with uh, yeah, shooting ranges and V8 engines. I'm not sure if you guys have witnessed this, if you go to any local ranges, but it seems like that's the, that's the pick, right? Someone's driving a truck or a Jeep in the parking lot.

Speaker B: I've spent a bunch of time at Excal actually, and that place I feel like, strangely enough has a bunch of sports cars. And other crazy stuff. Normally in the parking lot, but. Yeah.

Speaker A: Is there an Excal? Where, where are you?

Speaker B: Uh, so Denver's home for me, but, um, a lot of our team is Northern Virginia, kind of Ashburn area, so we spend a lot of time at Excal.

Speaker A: Yeah, that's the same one then. I didn't know if it was a franchise or not, but, uh, yeah, maybe my, my day, it was, uh, it was all V8s in the parking lot. Uh, cool. Let's pass it around to our guest Tyler. Uh, quick, uh, intro and your pairing.

Speaker C: Hey, Tyler Saltzman, CEO of EdgeRunner. I was gonna do something silly like French fries and ketchup, but now that we're doing cool things, I would say guns and watches. If you collect guns, I feel like you're also a watch guy and vice versa.

Speaker A: Nice. Any specific watches that, uh, you lean into? Any favorites?

Speaker C: I mean, my favorite is my Daytona. And then I, I'd say my favorite rifle is my scar 308. The uh, the 17, the scar. Heavy.

Speaker A: Sick. Good stuff. Yeah, much better than French fries and catchable though. Classic pairing. Uh, let's go pass uh, it around to Colton. Uh, quick intro and uh, and your pairing.

Speaker D: Yeah, Colton Alkerson, co founder and CEO at edgerunner. Um, love those pairings. Obviously endorse all those. But I would say I'm, I'm always trying to learn, um, and also trying to learn about kind of both the defense industry and startups and how you run a startup. Um, so I'm always listening to a podcast. Um, so I love listening to a podcast on a long run. Um, and it's either this podcast or uh, the second front podcast as well. Um, but then there's some other ones I like to.

Speaker A: Good stuff. Appreciate the plug too. Um, all right, yeah. Uh, bring us home. Tj.

Speaker B: Yeah, tj. Uh, Rao, chief revenue officer at Second Front. Uh, been with the company for about four years. Had a blast scaling technologies across national security. I'm, um, going to go different direction on the pairing. And Tim pulled a little inspiration from one of your previous episodes. Uh, but for all the folks out there that are also in the trenches of young children at home, I'm going to pair infants in remote controls. Um, no toy satisfies the, um, attention span of my 10 year old infant. Like just ah, TV remote control. We take the batteries out obviously, but, uh, she just absolutely loves it.

Speaker A: Yeah, it's pretty, uh, pretty affordable too. You save some money on the, uh, on the toy cost there. Uh, how old is your. Is your uh, your kid?

Speaker B: 10, uh, months.

Speaker A: Oh, nice. You're in the thick of it, man.

Speaker B: Good luck. Yeah, thank you.

Speaker A: Just one or, or two?

Speaker B: Yeah. First one.

Speaker A: Cool. Good stuff. All right, good pairings all around. Uh, let's, uh, let's go ahead and transition us into the, the heart of the discussion today. So, like I mentioned, we're gonna talk with the leadership from both edgerunner and second front, uh, around warfighter centric design, um, the infrastructure shift to on device AI. And then I like to cap with some more of a little open conversation around scaling in the dual use world. Um, but, uh, let's begin with building for the warfighter. So kind, um, of digging into the mission behind all of this. You're all building products that are, you know, dialed into, you know, uh, an end user, um, you know, uh, in military settings. The War Fighter, for example, you know, Tyler and Colton from the Edge Runner side, you know, what is war fighter centric design mean to you in practice? And how does it kind of shape your product strategy?

Speaker C: Yeah, so I'll start by, um, like, let's talk about the frontier models right now, like ChatGPT and Throbbix Clyde, and they're great. Um, ChatGPT created the market. It really show the magic of AI and how it can be useful. But I think the problem is it's not really solving for anything, even though it's really cool. Um, and then take it a step further. It's not personalized. So for example, you can't ask ChatGPT how to help you run an ambush or a move into contact mission or convoy operations. It's not going to know that. Um, another problem too is these big models have been trained on basically across the entire Internet. So there's lots of harmful biases and bad data in these models. And of course, if they hallucinate, it's much harder to guardrail them, um, to be safe and again, to be useful. And so what we're thinking is how do we build domain specific AIs that are smaller, that are specific to the warfighter, to your mos and then how do we have that live air gapped on a device so it never needs the Internet? And I think that's how we define the tactical edge. It's where the data resides. But tactically, with no Internet connectivity or any kind of Bluetooth or 5G.

Speaker A: Yeah, Colton, you want to add on to that too? Anything specific?

Speaker D: Yeah, I mean, I think it's really important to understand, you know, the situation that the warfighter is in today.

Speaker C: Right.

Speaker D: Most Situations are going to be in a denied or contested environment where you just don't have access to the cloud or you don't have great connectivity. Even if you talk to folks who are on a base here, you know, in the continental US they often don't even have access to their email because their network is so poor and spotty. So while I agree with Tyler that it's great that we're bringing a lot of these large frontier models into the dod, um, you know, there's going to be tons of use cases where those just can't support the warfighter. And that's why you have to have these models and these capabilities running on device locally, at the edge, where you don't need to have Internet connectivity. So I just think that's really important.

Speaker A: Yeah. Are there any specific moments where, you know, feedback from the field kind of directly changed your product roadmap or, you know, different design priorities?

Speaker C: Uh, I'll give some m anecdotal experience while I was, uh, serving overseas in 2017 in support of Operation Atlantic Resolve. We run convoy operations, and there was a particular instance where one of our trucks went down, and it was carrying an explosive ordinance. Now, the question was, can we repackage this explosive ordinance onto another truck with other kinds of explosive ordinance? Then the question is, if that truck were to get hit, would it exacerbate the explosion and take out the convoy, or would it be totally fine? You asked, uh, three different experts on that convoy. You get, yes, no, maybe. And me as the young lieutenant, I got to make the call. And of course, we rely on our chief warrant officers. They're the technical experts, but they're not always there. And so this could be a time where it's like, an AI would know this because it's trained on all this military doctrine. And then I could interact with it via natural language processing, like the conversation we're having right now. And I could say, hey, look, is this good to go or is it not? And now I can make the right decision immediately.

Speaker A: Yeah. Where time is, uh, of the essence. Right. Kind of breaking down some of that chain of command and empowering those folks to make those decisions quickly. I mean, seems like super key. Um, T.J. i, uh, want to ask you, I guess, from, like, the second front side, you know, you've. You've brought a ton of commercial solutions into national security missions. You know, what kind of stands out about Edge Runners approach to the warfighter use case and talk a little bit maybe about how this partnership makes sense. Yeah.

Speaker B: Uh, I mean, I think it's a refreshing approach to frankly bringing models to the warfighter. Like Colton touched on models that are very large, not necessarily fine tuned to specific data sets running on kind of your standard military network. Most of your soldiers, sailors, airmen, marines, they're not on a terminal at their computer when they're executing a mission. So the modality and the approach to the modality is first and foremost. And then on the partnership side was really exciting for us at second front is we've spent the last four years kind of, I'll say slogging through the trenches of bringing software onto those networks and we're pushing towards the edge as well of hey, yeah, we're going to need small models to run at the edge. We're going to have some challenges there. But from a managed hosting perspective we have to look at it pretty holistically where how do we deliver with one single platform layer in a secure and compliant way. Workloads to cloud, hosted networks, private cloud, hybrid cloud, on prem air gap, edge embedded weapon systems. Uh, there's a wide variety of hosting models that for companies like edgerunner to actually design with warfighter centric design processes, you have to be able to iterate in that capacity at speed. And if you like, that's our entire value proposition is to abstract the security compliance and infrastructure pain that comes with deploying kind of one to n environments. Um, so that's, I mean we were aligned from the start.

Speaker A: So Tyler or Colton, you guys are dual use, you know, you support commercial verticals as well. Did the company kind of start with the, you know, defense use case in mind first? Uh, or did it start in the commercial space first and was curious to hear, you know, what came first.

Speaker C: Yeah, it was Defense. Our, our first big caller customer, if you will, was, was the Space Force, Space Systems Command. And they, they really leaned in and so um, I actually met Lieutenant Deveren when he was an intern at aws when I was working at aws and you know, it's a small world, I go to stability and then after stability I leave. And Coltknife, we founded edgerunner and he's now a lieutenant. And the Space Force is like, hey look man, I'm looking at Generative A at the edge, the Space Force interested in this, let's work together. And then they brought us into Stanford where we had like a fun presentation, even a little bake off with what we could do versus what a Palantir or a GPT4 oh mini could do. And the objective was to uh, build um, it was an agent to help a new Guardian schedule satellites from the ground station via the proper SOPs. And we were able to do that living completely on a MacBook. And what was cool is the military didn't know that that existed. And so because they didn't know existed, they didn't know that they should be asking for it. And so that's really how it got started.

Speaker A: That's cool. What are the commercial applications that you see this technology as the best use case.

Speaker C: Cole, you want to take that?

Speaker D: Yeah, I mean, I think the defense use case for us is really clear, but frankly the value proposition of private, secure, uh, always connected AI is also really relevant to enterprises and especially regulated enterprises. So think of yourself as a, you know, metals and mining or manufacturing or transportation or logistics. All of those industries have, you know, pretty sensitive critical use cases where LLMs and AI can be applied. Um, and those instances are all, you know, those all need to be totally private and totally secure. So like our core product can be applied to each of those.

Speaker C: Frankly, um, we can't name the name of it, but a major movie studio, we actually took a mission planning agent and retrofitted it to write scripts. And so what they're doing is they're hedging against writers strike. And we're able to help them write, you know, beautiful scripts for, in their style for movies. And what's important is typically with a human, it would take them six months to write a script and you only get one version of it. With AI agents writing these scripts, you get a script in weeks to a month and you get many different revisions of that script. So we're able to help these studios innovate. I mean this can be used for gaming, it can be used for movie studios, um, healthcare, all that good stuff. Basically the bigger vision is the Jarvis, your Ironman, an AI assistant that lives with you, helps you. And I think, I think there's a future where even advertising becomes dead because these AIs will know what you want.

Speaker A: Mhm. Yeah. I want to ask you, tj, just because, you know, we'll dive deeper into some of the dual use scenarios, uh, as well. But you've seen a lot of, you know, dual use companies trying to build, you know, in military settings for the war fighter. What, what are some examples where you see companies kind of getting it wrong in that process? Like where, where are they kind of missing?

Speaker B: Oh, many, um, I think there, uh, there's a wide array of, I think where places where folks get it wrong. Um, at its simplest form, it Builds something that people want, that soldiers or an end user wants and is tied to a requirement that has funding like that is first and foremost kind of the easiest way to approach it. And then the second order of that is, are you building your product in a manner that will be consumable by this specific market that has its own esoteric rules and policies and procedures for how you deliver your product? And then two, do you have a go to market team that can go capture that business? Um, we've seen even recently in the last couple of weeks, there's some prolific venture capitalists in our space saying hey, we've got capital allocated to these businesses, but we're not getting contracts. And that's an execution challenge. Like there, there is plenty of capital and contracts and vehicles for great companies with great products to go capture, but the execution is very hard and it's not dissimilar to sales in highly regulated industries or in Fortune 500, but just product architecture and then real true grassroots go to market. I think a lot of folks get wrong these days.

Speaker A: Yeah, I mean the acquisitions piece is you know, a uh, massive, you know, hill to climb. It's, it's, it's clear like if you don't have somebody that understands the ins and outs of, of how to win that type of work, it can obviously be a, a detriment when you're on a, a tight Runway. Um, but from like product building, you know, we, we've partnered with a number of defense tech companies that look specifically when they're hiring up, you know, those early stage kind of like product owners, uh, if it's something that is, you know, if you're go to market is, you know, for tier one, tier two operators, for example, not having folks that have been in an operational seat in the past, as you seen that be detrimental or you know, building without, you know, maybe some former military folks in the product side of the business.

Speaker B: I think so I don't think it's uh, a silver bullet though. Um, like I think actually we, we over index to operational experience maybe a little bit too often in this market where I mean there's just a tremendous amount of technical expertise in this country that hasn't spent time with an operator or with kind of a mission first organization. So you can learn, I think what the mission set is and how folks operate and how they deploy technology today. I don't think that's the biggest hill to climb. Um, but it's certainly helpful. I think the key component is maybe indexing towards experiences that are analogous to the way technology is delivered in this ecosystem. So to Colton's earlier example about energy, oil and gas critical infrastructure. If you think about companies today that are operating, um, you know, nuclear power plants, they have laws and regulations that their software systems have to be air gap, they have to be on prem and you have techs out there actually installing software on site. So you think about a product like edgerunner and having AI at the edge that a soldier can use. It's trained on doctrine, trained on SOPs. That's perfectly analogous to a tech on site at a nuclear power plant or attack on site at an oil rig in Texas or off the coast. Um, so I think finding technical experience that's analogous to the delivery is maybe more important, at least in my mind than like, hey, you spent time with a tier one organization executing this specific over the horizon mission.

Speaker D: And one thing I just quickly add too is, you know, edgerunner is a dual use technology company, right? Like we're targeting defense for sure and that's our primary focus today. But we're also targeting and have enterprise customers. Um, and the DoD right now is very focused on commercial off the shelf technology and they want to bring the best technology that the commercial industry is using into the dod. Um, so I think it's really important that you actually have both DoD and non DoD customers because that'll help build a better product and that'll help serve your DoD customers even better.

Speaker A: Yeah, let's talk a little bit about I guess this infrastructure shift, um, and why edge AI is the future. Um, you know, we've kind of seen a lot of movement away from more centralized cloud models and you know, running towards AI at the edge. You kind of pointed out specifically in environments that are unreliable or sometimes just entirely unavailable. Um, Tyler, like what, what are some of the, you know, those real world constraints that are driving this kind of infrastructure shift and how is EdgeRunners platform addressing those?

Speaker C: Yeah, I'd say it's a few things. One, the best data strategy is one where you don't move your data. And I think we define the edge by where the data lives. That could be at a co location or it could be on a cell phone or a laptop or it could be in a Humvee. Um, so you have that aspect. But then if we look at like chatgpt or anthropics. M Claude, every time you're using it, you're paying by the token, by the word. That's very expensive and you're having incremental costs and it doesn't scale Linearly and so that's problematic. And then of course every time you're using it and you're asking or you're prompting those models, you're actually giving them data, you're giving them some of your IP. Like we saw the Samsung leak with OpenAI. You had a bunch of engineers refactor their code with GPT4. What does OpenAI get? All of Samsung's precious code. So we have those problems too of giving your data there. And then of course, um, there's bad actors that can intercept things. So we can eliminate all those risks and those incremental costs by having everything localized and the AI is brought to you.

Speaker A: What are some of uh, the trade offs though, like Colton maybe for you here deploying open source, um, LLMs more locally and you're kind of optimizing for those, those types of devices, uh, that maybe don't have as much horsepower. Like what are some of the trade offs that you, you kind of give up?

Speaker D: Yeah, I mean of course our, we have a research team, uh, you know, we've got multiple PhDs on staff, we've got a large engineering team and you know, what they're doing every day is figuring out how they can take, you know, the best of breed, uh, open source models, fine tune them on specific military data and then optimize them to get them to run on smaller edge devices without losing that performance. Right. So there is certainly some trade off you have to make between performance and where you're deploying. But what you can do is by tailoring the models and making them specific to an MLS or a task or making it just understand the army or the Navy or whatever the doctrine is very, very well and getting rid of all that other crap that, you know, these general models have, you can maintain that level of performance. So I think something that we're really trying to do is, is to not compromise on performance and not compromise on standards by making them very domain specific and making the models very specific to, you know, the given mos. Um, that being said, you know, a lot of people do like to use the larger general frontier models because they're using it for their homework or they're using it for their research project or what have you. And they do need to have a larger, more generalized data set of knowledge for those models. And that's great, but that's not what we're doing.

Speaker C: Right.

Speaker D: We're building solutions that are specific to the warfighter and specific to a given MOS or mission set.

Speaker C: How do you, there's one thing I want to call out too. One of the biggest ones is tokens per second. Everyone talks about that. So you can get like 300 tokens per second. But it doesn't matter because the average human reading Speed is only 6 to 10 tokens per second. So a big trade off is you don't have that super fast inference speeds but it doesn't matter if the human can't read that fast. So that's, that's probably the, the biggest trade off right there. But it, but again we don't believe that really matters because again you can't read that fast.

Speaker A: How do you guys prioritize, um, like which AI capabilities kind of make make the cut for edge deployment? Is there a, a technical call? Is it, you know, a UX thing? I'd love to hear how you prioritize.

Speaker C: You know what we're doing is we're constantly looking at uh, you know, the Colton's point, the best open source models, you know, whether it's a meta model, a uh, Google model or from Mistral. But then what we do is we're, we're thinking about, all right, which benchmarks are these different models good at? And then with our data set, we built a 30 billion token data set of military doctrine. And then what we do is we run different benchmarks based on these different models. And so we'll have a model that's 24 billion parameters, a uh, 12B and a 4B. And we're, we're working on a 1.6 B and then it's this suite of models that's constantly changing because every three months there's a new model out. And so the shelf life of these models is three to six months. So we have to keep constantly looking at what's the latest and greatest model. But that doesn't mean it's better. But if these models are like a year old, like llama 3, 18B, they're going to be completely outdated. So there's a balance there of like, all right, what's the latest and greatest? But at the same time what's again what's creating the best product? Because again models aren't products. You um, brought up the UI UX component. Like customers want an entire experience, they don't just want open source models. And so that's another thing that we're solving for.

Speaker A: Tj, what about uh, from your all side, you know, you're seeing a lot of these, you know, more and more systems kind of moving to the edge. How does that shift? Second front's kind of approach to security and, um, compliance.

Speaker B: Um, yeah, I think you have to meet the customer, where they are and what their posture is on security and compliance. Um, there's even a perspective, I think today that, you know, Air Gap on prem is more secure than cloud workloads. And that's not necessarily true in all cases. Um, our perspective is that we should be pushing the limits on how we're deploying AI, how we're securing AI, uh, but in a secure and compliant way. I think the unfortunate reality of most of our software systems today is that they're compliant but not secure. And a lot of products are secure but not compliant. And ultimately that just slows down how quickly companies like edgerunner are iterating on their product with the warfighter. Um, so that's our substrate, is how do we create speed, um, in kind of the face of uncertainty around what does security and compliance actually really look like for a lot of these models, which is still undefined to a certain degree.

Speaker C: And another reason why I'm really excited about our partnership with Second Front is for those that don't know your ATO, your, um, authority to operate, there's like over 800 different security checks. It's crazy. And Second Front streamlining all this for us. So we ask about, like, again, what are, what are some of the, you know, difficult challenges of, of building a startup for the, for the DOD? It's getting that ATO and understanding that grueling process of 800 plus security checks. And so folks like, you know, TJ, they help us get it done, which is super helpful. And then of course, they can help us orchestrate this everywhere.

Speaker A: How did that partnership come about? I'm always curious on, uh, was it like a connection through, uh, a venture partner, or did you all know each other just through the defense tech ecosystem?

Speaker C: Yeah, they were actually on Colton's radar. First. Colton, you want to dive into that?

Speaker D: Yeah, well, obviously we knew of Second Front, um, just through the ecosystem, given how successful you all have been. Um, but yeah, we actually were connected, uh, I believe through Salesforce Ventures together, um, which is just sort of the power of this ecosystem and investors sometimes can actually provide value.

Speaker A: That's cool. Uh, I want to just do a quick zoom out, um, around scaling in the dual use world. We talk a lot about it, uh, on the pod and obviously love to hear that firsthand perspective from founders that are doing exactly that. So, you know, for, for Tyler and Colton specifically, you know, what's been, what's been surprising or difficult, uh, about scaling a dual use Company, you know, across both DoD and commercial sectors.

Speaker C: You know, Colton made a good point that the DoD wants commercial off the shelf software. That's proven in the commercial world. But one of the biggest challenges is you need to stay focused. And I don't think there's too many startups that fail because they're too focused on something. I think the problem is startups will try to do too much at once and then they'll fail. And we're trying to find that balance. I'd say that's a tough balance. How much of our time is allocated towards commercial versus DOD? I'm thinking the 20, 80% rule. Um, but of course it's fluid.

Speaker A: Um, Colton, any perspective that you've seen?

Speaker D: Well, I think that's spot on what Tyler just said. I think the other interesting thing, and this is both challenging but also helpful to us, which is generative AI right now is so new, right? There's a lot of people who are AI engineers 10 years ago and they're out of date now for a lot of what we're seeing today. Or you could be someone new, like, um, a young kid out of college or even high school and you could be an expert on this stuff that, you know, vastly outperform someone who's been doing this for a decade. So I think something that we've also been trying to adapt to is, hey, how do we constantly stay at the cutting edge of this technology, both on just sort of our knowledge base side, but also on the hiring side. Um, and we've been able to hire some people into our company who, you know, they're very young. They're very young and they don't have, you know, 20 years of experience in AI. But they are frankly far superior to many people who've been in the industry for a long time. Um, so I think it's been sort of both, um, it's actually been very good, uh, for us as a startup trying to sort of take on, uh, some of the larger, well, um, funded incumbents because we have that opportunity.

Speaker C: A, uh, funny joke is, you know, when we're at stability, you'd see kids with like, discord names of like Deez Nuts outperforming, uh, Google DeepMind engineers never gets old PhDs. It's just they're in their grandma's basement and they're outperforming the seasoned vets from DeepMind or Meta. And so it's such a crazy world. It's like, how do you find, how do you find that talent? Cause you can't look at traditional resumes anymore.

Speaker A: Yeah, when you build out the teams, um, you know, I could see obviously the use cases on the sales teams having, you know, more of that domain expertise. But within like product and engineering, do you, do you, is there any split to that? Is it all, you know, built, you know, together, uh, in support of all customers or how do you kind of build out those teams?

Speaker C: It's sort of like, you know, with traditional SaaS, you have your engineering team and then you have your product teams. But for what we're building, it's almost like the product is research and engineering fusing together and, but you know, so that's, and that's the product. It's like that's, I'd say that's, that's, that's a big difference here for like new product leaders is, you know, how do you, how do you really manage between research and engineering to make sure that you know, you, it's like you have your theoretical research with models, but then how do we apply it? And that's the applied engineering part. And then you have like, then how do you put it all together and like a usable, you know, product that's intuitive. And so I'd say that's, that's been, you know, uh, challenging but super interesting. And we've, we've learned a lot running that process.

Speaker A: Yeah, tj, you kind of touched on it you know, a little bit earlier, but any common, you know, traps, uh, or misconceptions that you kind of see when it comes to, you know, selling into defense as a dual use startup?

Speaker B: Um, yeah, many traps I think, and we've all fell for them. You, uh, spend time with great customers that don't have funding on the back end or you don't have a vehicle tied to actually draw scale and growth to that customer or that agency. Um, I think my point on maybe scaling uh, in this market, I think scaling a company anywhere, particularly hyperscale, is tremendously hard. Um, but if we're all doing our jobs right as founders or operators, you really have a different business and a different product every six to nine months. And that's an incredibly hard space to lead and manage and hire and grow into. Um, so we've made many, many mistakes in the past, but just indexing on like the intangibles of the, the talent that you bring into your company and focus towards mission, I think is, is the best way that you can go after it.

Speaker A: Mhm. Are there any like ah, specific tools or marketplaces that you've seen, you know, super Advantageous, like, like trade winds or anything specifically that you've, you've seen kind of move that needle to help towards, you know, specific wins or maybe insights.

Speaker B: Yeah, I think they've moved the needle maybe on the macro perspective of getting the DoD broadly more comfortable with marketplace type transactions and going through whether it's trade winds or a, uh, CSO kind of from DIU or even transacting through like an AWS marketplace, um, you know, two or three years ago, I don't think we saw any transactions going through there. So it's moved to the inertia. But from our second front perspective or edgerunner perspective, I think you just have to think of it as one piece of a larger puzzle in your acquisitions and sales strategy that you need to have vehicles and transaction pathways in place for a myriad of customers. You are going to have organizations and agencies that prefer to buy through a value added reseller and they have their vehicle set up and their contracting shop is tied in and that's your fastest pathway. You'll have contracting shops that know how to move money to trade winds and execute a contract through that. You're going to have pathways that have to go through a, uh, traditional RFI and RFP process. And you would just want to make sure your team has kind of the skill set and tools to go execute wherever that pathway your customer leads you to.

Speaker A: Tyler, anything specifically that you know, you found to your advantage when, you know, kind of doing more of the sales. On the defense side, you mentioned Space Force. I didn't know if it was any specific, you know, stratified tac fire, spaceworks type of engagement. Anything that, that you found is advantage. Ah, in those early stages, I'd say

Speaker C: the quicker, the quicker you can get to the pos, the better. Um, I don't want, I don't want to complain, but like the server process, I feel like it's broken. It almost seems like it's already earmarked or rigged for someone somewhere, somehow, probably through the PEO office. So I'd say, you know, get with these POs, build a relationship with them, with him or her, uh, and show them what you can do. And just because we're obsessed and don't get caught up chasing sivers and baas and all that stuff, build your product and try to meet with these, try to meet with your customers as soon as you can and get these stakeholders to lean in and help you out.

Speaker A: Yeah, yeah, it's good feedback. We've talked to a lot of folks that have been just trying to uh, dissect the silver process and how so many companies get fizzle out in that phase two and never get to a program of record or, um, SHIELD Capital was on previously and, and really had some sound insight in terms of, like, just not getting. Not getting lost in, like, non dilutive funding and just thinking that's the pathway. Right. Because I think that's. It's easy to do so and, and, uh, appealing. It's non dilutive, but it's also, uh, not only going to take you so far, um, but, uh, I want to wrap, I guess, with a little bit more around, you know, where you guys are. Edgerunner, um, as a company. So maybe just a couple of quick hits in terms of, um, you know, headcount, you know, where. Where are you guys based out of, um, you know, funding to this point? I'd love to hear just a little bit of that.

Speaker C: Yeah, so we. We were founded back in February, so we're. We're pretty new. I'd say what I'm most proud of is we already have a live deployment with SOCOM overseas in Asia. Uh, and. And they're using us and you know, for. For real missions. It's not just a pilot. And so for us being as young as we are to. To be there, definitely, um, part of my team, we're 21, full time, uh, about to be 22. We have our chief Science officer joining us, um, I think next week. Colton.

Speaker D: Two weeks.

Speaker C: Two weeks. And then, yeah, we've, um. I mean, we raised 17 and a half billion total. We did a five and a half million seed. 12 million a. And, um, you know, we're still young, but, you know, we're. We're getting after it.

Speaker A: Yeah. What's on the horizon, Colton? What's. What's, uh, on the next, you know, one three year plan?

Speaker D: Yeah, I mean, so we just released, um, um, a DOD focused beta of our product. So go to edgerunnerai.com, uh, and if you're in the DOD, you can sign up and, uh, actually download our product and test it yourself. So we'd love for you to kick the tires on it. Um, we're going to keep rolling out different pilot deployments, uh, with different customers. Like Tyler said, we've got a deployment right now that's live and active in Asia with socom. We've, uh, got some other deployments coming online that we'll talk about later this summer and this fall, which we're really excited about. Um, we'll be releasing an update of our product here in about two Weeks, uh, with a new model, uh, that'll actually have additional capabilities and allow us to support even more people across the dod. Uh, but really what we're already starting to do is look at other capabilities for AI at the edge. Um, so given what's going on in robotics and in drones, there's a ton of use cases for computer vision models and running models on drones at the edge, also in denied and contested environments. Um, so we'll be working on that quite a bit, uh, as well, um, as well as some other things related to AI at the edge.

Speaker C: And we're doing that through a crada, a cooperative research and development agreement with the Army Research Lab for the robotics piece. And then we have a cradle with AFRL to build, uh, AFSC Specific agents, or MOS Specific Agents.

Speaker A: Very cool. Tj, I'll give, uh, you the floor, too. It's what's, uh, new and exciting on the second front, the near term horizon.

Speaker B: Yeah, I mean, uh, on our horizon, what's new and exciting is kind of the same thing we've been doing for the past four years, and that's deploying workloads, uh, on behalf of our mission partners to the warfighter. Um, so our mantra is deploy the platform to new places to include Europe, uh, apac, our allies and partners, and deploy more workloads out to the warfighter in a modern, secure, compliant way.

Speaker A: Cool. Nice. Um, I think that's a wrap for the main conversation. I do like to close with, uh, a little bit of rapid fire Q and A call, um, it the five second scramble. So we'll just kind of go around the table, do five questions for each of you. Some business, some personal. Uh, Colton, I'm going to lead off with you. Um, so if Edgerunner were an animal, what animal would it be?

Speaker D: Uh, an eagle.

Speaker A: What's one piece of tech equipment aside from your phone that you can't live without?

Speaker D: Uh, tech, I would say my AirPods. It's adjacent, but I'm always on the phone or listening to a podcast, like I said.

Speaker A: Yeah. What are some tech roles that you guys are hiring for over the next three to six months?

Speaker D: Hiring for forward deployed engineers. Always looking for best of breed researchers and engineers. Uh, and then probably later this year some more go to market.

Speaker A: Folks, what was your first car?

Speaker D: Uh, my first car was a Volvo S60.

Speaker A: Nice.

Speaker D: From my, uh, parents hand, uh, me down. To be clear.

Speaker A: What was your dream job as a kid?

Speaker D: Uh, my dream job as a kid. Honestly, I always wanted to, um, serve in government, either as an elected Official or run for president someday. Um, I worked in government, and now I know much better and would never want to run for congress or elected office ever. Um, but I'm happy to serve in other ways.

Speaker A: Good stuff. Tyler, I'm going to jump over to you. Uh, what's a core company value that you find unique about the edgerunner culture?

Speaker C: We're direct, we encourage debate, and we like to say, execute with violence, Move quickly.

Speaker A: Nice. What was your call sign in the army?

Speaker C: Salty Snipes.

Speaker A: Nice. Uh, what's a charity or corporate philanthropy that's near and dear to you, Big

Speaker C: brother for foster kids? I was. I was one.

Speaker A: Cool. What was your first job?

Speaker C: I was a. I was a pool boy when I was a. When I was a high school kid.

Speaker A: Nice. Now we know the headshot, uh, image we're gonna need for your episode here.

Speaker B: Yes.

Speaker A: What was, uh.

Speaker D: Oh, no.

Speaker A: What is your. Your favorite superhero?

Speaker C: Magneto.

Speaker A: Cool.

Speaker C: He's got that anti hero aspect. So misunderstood. I dig it.

Speaker A: Yeah. Nice. All right, uh, tj, we'll close out with you. Um, if Second Front were an animal, what animal would it be?

Speaker B: Uh, elephant.

Speaker A: What's your favorite part about the culture at Second Front?

Speaker D: Mission.

Speaker B: Uh, first, for sure. Um, we're an enabler for all of our customers. We are not the actual application or the workload. Uh, so, Mission, uh, go ahead and

Speaker A: plug some of the types of tech roles that you guys are hiring for in the next three to six months.

Speaker B: Yeah, a ton. Uh, we got developer roles. Uh, platform engineer, product manager, security engineer, network engineer. Uh, a whole suite.

Speaker A: Sweet. Uh, and last, uh, two. Uh, favorite charity or corporate philanthropy that's near and dear to you?

Speaker B: Favorite, uh, charity is the Station Foundation. Um, they operate kind, uh, of a wide array of on site, um, kind of courses for special operators and tier one families coming out of the community.

Speaker A: Very cool. And last one. What is one thing that you hope I never replaces?

Speaker B: Oh, I hope it doesn't replace, um, good taste for food and drinks.

Speaker A: Yeah. Well said. All right, that's a wrap, guys. Appreciate you spending time with us on the podcast. Colton, Tyler and tj, thanks for joining.

Speaker C: Thank you.

Speaker B: And thanks for having us, you guys.

Speaker A: Sam,