Next-Gen Defense: AI on the Frontlines with Thomas Robinson, COO Domino Data Lab

Thomas Robinson Keywords defense tech, national security, AI integration, commercial success, procurement challenges, DIU partnership, Project Ammo, legacy systems, AI governance, talent acquisition, public-private partnership, venture capital, dual-use technologies, domain expertise, mission requirements Lauren Bedula 0:00 Welcome back to Building the Base. Lauren Bedula and Hondo Geurts here recording live from the Reagan National Defense Forum in Simi Valley. Pumped to be out here. There's a lot of positive energy around defense tech and modernization related issues. We're sitting here with Thomas Robinson, aka T Rob who is the Chief Operating Officer of Domino Data Lab, but has a really interesting background in the finance world, and is super patriotic and eager to be doing everything they can for the defense and national security apparatus on top of successful commercial work. So T Rob, thanks for joining us. We're excited to dig in. Thomas Robinson 0:34 Thank you for having me. Hondo Geurts 0:35 So T Rob, give us your background? What's your origin story of going from successful businessman to now in defense tech, out here at Reagan. Thomas Robinson 0:43 I'll go back a little bit further. I'm a nerd by training. I studied math and computer science in college, and then after that, as Lauren alluded to, I went to go work for a large hedge fund for about 10 years, building enterprise technology, which you know, it was building and making critical finance decisions, and then joined Domino after that, about a year after after Domino was founded, and have had a variety of different roles, but most recently, I'm focused on building the go to market for our business and so marketing through to delivery of our solutions and software. Lauren Bedula 1:22 So what is it that drives you to national security and defense? Thomas Robinson 1:27 Yeah, well, you alluded to there's a patriotism first. And you know, I've been, I think I mentioned to you yesterday, I recently became a US citizen, and I have been blessed by this country, and I feel there's a need to pay it back, and the technology that we have built in the private sector, I think, has a fantastic match. And, you know, we're here to serve. And I think most importantly, there's a moment now, I think, catalyzed both by the advent of AI and I think changes in the way the DoD is approaching Silicon Valley and the way Silicon Valley is approaching the DoD, it's mutual. That makes an amazing moment for us to have an impact on refreshing the industrial base, and, in particular, supporting the mission of the warfighter. Hondo Geurts 2:28 So why don't you give us, as we kind of dig into this, a bit kind of what's the founding principles of the product, and where you've seen commercial success, and how does that translate into viewing the defense tech market? Thomas Robinson 2:43 Yeah, absolutely. So we're a technology company building software and services as integration is critical to many of our customers, and the core philosophy of our product is to enable advanced domain experts to build advanced models, data science, ML, computer vision, deep learning, generative AI, the whole gamut, and do that in a way that has fast time to value, so getting the model to market as quickly as possible, but building trust along the way. And where we've excelled in the private sector is working with large enterprises that have models at the core of their business. What does it mean to have a model at the core of your business? It means that your product development, your critical business decisions, are built on algorithms and models that are making those decisions for you, not always autonomously, sometimes autonomously, but by way of example, the few areas where we focus, financial services. So we work with globally systemically important banks who build their risk models. You can imagine billions, 10s of billions, maybe even trillions, of flows of capital, and having models that assess the risk that is critically important, and so you need time to value, to respond to market trends, and you need trust. We also have done a lot of work in pharmaceuticals. We have 10 of the top 20 global pharmaceuticals, our customers who do everything from new product development to speeding clinical trials with our product, and again, somewhere where human health is critical, and trust is important, but time to value is equally as important. We had a customer tell us every day we can speed up a clinical trial. We save ten million. Wow, right? Like, that's pretty amazing. Lauren Bedula 4:34 Wow. Now talk tell us what you see in terms of differences dealing on the commercial side and then selling into the US government, any kind of takeaways in terms of the pros, cons of each? Thomas Robinson 4:45 I'm going to say, first of all, there are a lot of similarities, actually. I mean, I think this is one of the myths that is sort of perpetrated, that things need to be so radically different. And I think it's on them to sort of dispel that notion, a little bit like everybody we talked to. I was just, I was just speaking with somebody from cyber command who was saying that they're working on delivering AI technology on 90 day sprints. And I'll tell you that can be faster than some of our commercial customers, sometimes and and I think a lot of of what people have in their minds is that DoD is slow. I don't think that's always the case. I will say procurement is a challenge. I said before that, you know, I believe that DoD is making strides to move closer to Silicon Valley. I think Silicon Valley has come a little way. Still has more to go. And, you know, at the end of the day, the way procurement works in the DoD doesn't always match what a venture capitalist is looking for in terms of annual, recurring revenue and predictable revenue for their company. And so there are some gaps there that I think are being worked on. We've had the blessing of working with DIU, who I think has really made fantastic strides in changing the way things are done. And their magical moment of realizing the OTA would be a way to procure is a great start. But I think there's more work to be done. Hondo Geurts 6:15 As a former finance guy, I'm interested in your view, you know, [in how] valuation is viewed. I think one of the challenges under traditional mechanisms, when looking at the DoD, it's not a traditional customer, it's got a different buy cycle. It buys more when things are bad and less when things are good. Do you sense that the financial community is starting to pull together a rubric of financial measures for defense tech that's related to but maybe a little bit more unique, or at least getting more educated on the defense tech market when you're looking at valuing a company? Thomas Robinson 6:54 Yeah, I think we're in early innings of that. I'm not sure we've quite figured it out from a venture capital perspective, specifically, I'm still still not sure there's entirely a new mechanism what, what has changed, though, is the appetite to work with the DOD and and almost a bit of a cultural shift. I mean, I think we talk about this time and time again, there's more positivity about engaging in defense and that that has changed a lot from 10 years ago. I lived in San Francisco 10 years ago. It was very, very different at that point. And so, you know, the money and the mechanisms will follow. I think people are trying to figure it out. There are larger pools of capital going into the organizations. I think that that's primarily the first thing you need when we always talk about the valley of death, right? And like DoD is trying to shorten that valley of death. More money from venture capital helps with that as well. And I think, you know, there'll be a there'll be a snowball effect as as folks realize, there's a way to work here, I'm just not sure that it's going to shift overnight from looking at product led growth and annual recurring revenue to large contracts on multi year time horizons. Lauren Bedula 8:14 So let's go back. You mentioned you're doing some work with DIU. From what I understand, there's been just great progress in terms of a partnership between some disruptive tech companies access to the customer, sounds like a good news story to tell. Can you tell our listeners a little bit about Project Ammo? Thomas Robinson 8:31 Yeah, absolutely. So going back a little bit. We'd had some early commercial success with some defense contractors and some fed civilian agencies, and we made a bet that we wanted to go focus on public sector in earnest, and have built a team around that. But one of the catalyzing moments was working with DIU. So the Navy was looking for technology to help speed the time to model deployment and integrate some critical issues around mine countermeasures, so unmanned underwater vehicles that are gathering a variety of signals, information, computer vision, integrating that with models to detect where mines are located. That's critically important for fleet movements and for detecting hazards. And the thing that they wanted to do was speed that up. Their existing process took somewhere between six and 12 months to do that. So they worked with DIU. DIU put out a notice to identify vendors in the space mlops, I think was what they what they particularly said, and something that we focus on. We responded to that. Got a initial OTA via the DIU, and began to work with the Navy directly and helping integrate other venture under other vendors in AI in the space. And over time, we achieved a success memo, which is a massive move forward for us, maybe that's something the VC community can look to for, for what it means to be on a good path. Certainly, our board was happy about that, and you know, from there we've really been accelerating the work we're doing, the outcomes that we help them deliver, are taking that model development time and retuning time from six months down to two weeks. So think about the impact of that, right? If you have unmanned underwater vehicles that you need to move from one theater to another, and suddenly the computer imagery changes, the conditions of the signals change, and you need to retrain a model. Two weeks is far better than six months. Six months basically means you're not able to use that technology, right? So you know that that is critically important, and as I think, the ultimate success that we that we help provide there. Hondo Geurts 10:52 So one of the challenges with, I think, big corporations or the DoD is a lot of legacy systems complex don't always talk to each other, built on a bunch of different technologies. How have you guys tackled integrating with legacy systems, and what lessons might you have for other folks who may have a piece of a technology or an algorithm on how to partner to get an integrated solution to the customer? Thomas Robinson 11:20 Yeah, I mean, I think that AI makes that challenge doubly hard, because it's not just that you need to integrate a disparate set of technologies, but like I was saying with the Navy case, what you need is a rapidly iterating environment. It is not that you can build a weapon system, go through Test and Evaluation and then have it on the market for 10, 20, 30 years, you are literally needing to redevelop that model on a weekly or daily basis. And so I think that further raises the stakes of what's required for integration. The role we took with the Navy is, I think, what we're going to call being an AI prime. So there's a different need for a different skill set around integration. It's a lot more about the data, moving it incredibly quickly and then getting the models back deployed, but also monitoring and governance along the way. And so that's a different skill set than maybe a traditional systems integration I think obviously there are large components of that, but you also need folks who have data science and AI know how, who understand what data is required. Is this the right data set? Do we need to go procure a different data set somehow? Where are we going to go to do that? How are we going to integrate yet another system? How are we going to map and go across the large data set? So you know this, this challenge is larger, and I think it calls for new capabilities, and we're calling that AI prime. Lauren Bedula 13:03 Awesome. I want to pivot to talk about talent and what you're seeing in the workforce. We often say, Well, we focus on technology. Ultimately, it's a human endeavor. What are you seeing as you're growing this company? Is there a lot of energy to work on these issues? Or what's your take there? Thomas Robinson 13:18 Yeah. I mean, I think there's been a cultural shift. Like I said, I see a lot of folks having more passion about working on these issues, or at the very least less apprehension and, you know, we've along the way, as we made the endeavor to dive into, the investment to work with, with the DoD further, I think people would often ask me the question of, how's your workforce going to react? And the way we dealt with that was very straightforward. At our company All Hands and in our slack and all the technologies we use, we'd highlight the work we're doing and let people react right? And we want the people in our organization who react to that with gusto and pride and want to dig in more, and that sorts itself out. And I think that sorting is going to happen. You can go work at plenty of companies that don't want anything to do with with with national defense, but I think the the overarching cultural changes is people are more more excited about that. And I think they understand, I think they understand the implications, you know, we're hearing a lot about it here, right, that American supremacy needs to be reinforced. And I think people see that democracy and capitalism, you know, we help export and uphold those values across the world through strong national defense. And I think that that's a lot different than thinking about it as I'm building a weapon, right? It's a very different thing. And I think people are seeing that. Hondo Geurts 14:59 Yeah, what's your take on, you know, the debate, you see it in Europe right now about AI governance and policy, and, you know, you see both ends of the Wild Wild West, kind of AI is going to take over the world to it's got to be closely controlled. And my sense is there's probably a nice middle ground in there. What are you seeing? You're in these highly regulated commercial companies. What are you seeing there? And what lessons maybe for folks in the industry or government as they think about policy and governance on AI? Thomas Robinson 15:35 We don't think about the policy too much. I would say, I think the, you know, the policy is fine at the end of the day, the sorts of problems we work on require trust. I said that up front. And so every organization we work with, if they have decisions on the line being made by those models, they want governance. And you know, they want this balance right? They want the time to value, but they want the they want the governance. They want something that can be trusted. And so I don't think that the prime motivator of that is policy or regulation. I think it comes from the criticality of the problems you solve. So what I'd hope our leaders think about with respect to the policy is dividing up the use cases into those of importance and those of not so much importance or impact, if you will, right? So using a chat bot to generate an email and make yourself faster, I don't think we need too much policy around that. I don't think that's going to radicalize America in some way, right? Like I'm not too worried about that AI that is helping inform a battlefield decision, I better have some governance on it. And not because, you know, there's, there's a general reason for that, but because, if you want the trust of the war fighter to rely upon the AI, then it has to be right, and so all of our motivation for governance comes from that, and that's driven by the private sector as well, in developing drugs and in their financial models, in their insurance pricing models, all of those exactly the same thing. I think the best way it's going to happen is public private partnership, development of regulations that that make sense, that are developed with industry together, based on the actual import of the models that are created. Hondo Geurts 17:27 So I think that the takeaway, and that's a really powerful way of thinking of it, treat it if you're a government person, treat it as a mission requirement, not as a policy overlay. And if you state it that way, as a way to make sure you get that capability you're looking for. Thomas Robinson 17:42 Yeah, and in the transition, you know, like the executive order came out, and a lot of folks are think, Oh, well, the executive order may not be there anymore. Okay, that doesn't matter. The decisions are the same the mission, the missions are going to be the same under the hood, and the trust needs to be there. Lauren Bedula 17:59 Now, you talked a little bit about interacting with your board and your investors. A lot of our listeners are entrepreneurs, founders working on the private sector side. Of course, in addition to our government listeners. Do you have any advice to founders and entrepreneurs who are raising capital with an eye towards defense tech applications? Thomas Robinson 18:19 Yeah. I mean, I think a couple of things. One, don't go into this lightly. I'd say, first, do your research and do your homework to understand the reality of how it's going to go. Because the the valley of death is real to some degree, and at some point you're going to be in a discussion where it's like, Ah, this is not quite going the way we expected. And, you know, and you're gonna get some flack for that. So learning the realities of how it can go ahead of time, and planning against those thinking about it on a three year cycle and not a one year cycle, right? And thinking about the contingencies, and thinking about how much investment you want to make of your capital before you see the results is critically important. So I would say, you know, go find advice from people who have been there, and then expose that education to your board and investors, as well as you put together a plan a rosy picture. This is not the place for a rosy picture, right? Like there is a lot of hype in fundraising, a lot of hype in startups that's all about to the moon growth, you know, hyper growth, all that. I think the current economic environment has tempered some of that, which is good. There's a lot more focus on profitability and that sort of thing. But that makes the specter of the valley of death even more critical to deal with. So I think I'd first say, educate yourself. Find people who have been there and done that, use that that education, or even use those folks to help educate board and investors as well. Hondo Geurts 19:53 One of the things I often hear from government teammates is we, particularly in the DoD, we think our problems are all unique and bespoke, and that, you know, general commercial technologies, that they need to be, you know, featured and tuned and different. One, do you see that attitude still coming from government customers, at the detriment of speed or efficiency and two, do you really believe, as we go and continue to, I would say, blow up this commercial AI, do you believe the military problems will remain as unique as they are, or do you think over AI can help make what used to be unique problems more generalized? Thomas Robinson 20:42 Yeah, I'd say a couple of things about that. First, I'd say there is certainly desire, I think, to bring in those technologies, and a recognition that that dual use technologies are going to be going to be critical. Now, of course, on the tail end, there are always requirements that come up and the process you have to go through. And so there's always, there's always lingering challenges, but that appetite is there, and I think one thing that has helped really drive that change is the the advent of Chat GPT and generative AI. We had Neil deGrasse Tyson at our conference last year, and he made a really amazing point that I that I like to repeat. He said, You know, I've been dealing with AI. I'm not going to do my Neil deGrasse Tyson voice. I've been dealing with AI and and computers taking over my my role for my entire career, like, you know, 30, 40 years, I've been dealing with computers doing all this stuff. And suddenly it comes for, it comes for the liberal arts majors and people who do writing, and now everybody's all worried, right? But, but you know, what that's done is really exposed this at the forefront and made it palpable the need to go invest. And I think that's been reflected. The thing I would say as well is, I think the problems that the private sector solves are very, very applicable to what the defense department needs to do. However, if you are looking to take an off the shelf AI model and and plop it into a decision making process in the DoD, that's not going to work. And where we come from, our core philosophy has always been, you need domain experts empowered to rapidly build AI in a trustworthy sense. You have to have the domain expertise behind it. And so what we think translates very well is the process, the technology, the integration, but not the particular model. So we're we're not in the business of taking a pharma company model and moving it over to something in the DOD. We're in the business of building that efficient process and helping bring bring that across. I think one of the challenges for maybe our government buy teams is it's hard for them to discern all the different pitch decks, everything everybody's put AI on the end of any product they have. Do you have like, two or three questions, if you were a government procurement person trying to see if AI could help fit something? What would be like your two or three diligence questions that they should they should ask anybody coming to pitch them. I think I'd first ask to speak with a customer. I mean, at the end of the day, that that's the litmus, that's that's the touchstone that allows you to tell whether something has worked or not. So show me a customer where you have deployed your technology, AI enabled and the results that they got. And don't do that with the vendor. Do that one on one, and really learn and listen. And if, if the vendor is ready to line up and give you 2, 3, 4, 5, of those. Fantastic. If the vendor is going to struggle to do that, you know, that's an issue. I think the second thing I'd say is, you know, may not always have this capability, but bring somebody technical to bear. We've been massively impressed by the technical depth of the folks in the DOD, and they're there, and they understand who to how to pierce through the marketing veil by probing in and understanding how the technology actually works and what it actually does. And so those would be my two recommendations, references and bring technology experts to probe. Lauren Bedula 24:26 I have a quick last question, because we have another shared interest in mind, which are 16-17 month olds. Except I have one. You have two. Any advice to the listeners, how do you stay fresh and balance personal and professional goals? Hondo Geurts 24:39 Gosh, I mean, don't come to Reagan on a weekend! [laughs] Thomas Robinson 24:42 Yeah, my wife's gonna listen to this and slaughter me. She's alone with both of them. I just got a text of a toy in our dog water bowl. I texted a picture of the F-117. She texted that back, and she's like, this is how my day's going. So, I mean, yeah, I'm first time father, and I think having kids has opened my horizon to the future and like this, this is patriotic again, like I think about what the world's going to be like in 45 years, when I'm not around and they're going in, and it's, it's like giving me hope and motivation to go build and work on things that aren't just sort of as selfish and I don't know I mean that that doesn't help with, you know, 2am screaming and waking each other up, but it, but it does help me stay focused on what's important. Lauren Bedula 25:36 How cool. I'm glad I asked that. That's an awesome note to close on. So T Rob, thanks so much for coming on. Know you have a busy day. We appreciate you sharing all this insight. Thomas Robinson 17:58 Thank you for having me. Transcribed by https://otter.ai