In each of our podcasts, we ask top hardware entrepreneurs the same 10 questions to better understand the challenges and best practices in starting a hardware company. In Season 2 Episode 7, Lemnos’s Eric Klein speaks with Josh Ouellette, co-founder and CEO, and Randy Shults, another co-founder of RavenOps, a Lemnos portfolio company. RavenOps takes inspiration from nature to build cooperative swarms of simple robots that outperform individual units in complex environments.

Josh: It really goes back to undergrad for me. My focus at MIT was on aerospace control systems. Which I really liked because I’ve always liked software, but at the same time there’s this kind of magical moment that happens in controls, where I’ve got this lifeless block of motors, actuators, and sensors. Then you throw the switch, and the control system kicks in. As a company, we’re focused on swarm robotics, of multiple units working together, which we call “heterogeneous multi-unit robotic systems.” That same thing takes place again, but now at the systems level. So it’s not just one robot coming to life; it’s all of these robots in a dynamic system.

Randy: This was the first time in hardware. One of the scariest tasks we needed to tackle was trying to form a hardware startup. But that’s actually where swarm really plays into this. Instead of building a much more complex mechanical engineering system, we’re going all the way to a bayesian dynamic system. With swarm, we’re able to use off-the-shelf robotics, build in our own sensors, and turn this back into a software dynamics problem of how to interact between the different robots. So we’re much more into what traditionally has been our sweet spot.

Josh: I spent 11 and a half years prior to doing this in the Air Force, flying V-22 Ospreys and doing special operations in that aircraft. You have this weird relationship with a flight control computer—when you as the pilot are asking the airplane to do something, and it will either do it or not do it based on its own conditions and its own interpretation. And so, when our drones are flying around and they’re making all these autonomous decisions, it’s always in the back of my head, “What would it be like if I was actually sitting in the middle of that thing?” Because that’s the world I lived in.

Josh: When we were looking at oil and gas and some of these heavier industrial applications, we had this rule: we wanted to switch from us pushing toward customers, to receiving pulls from customers. We weren’t originally looking at this internal space, where we actually go inside of their equipment and do these inside-out inspections. Customers were telling us that. And what we noticed was, when we finally picked up on that signal, and pitched that back to customers, they said, “I need to get you to talk to so-and-so and so-and-so, because this is big.”

Randy: Through our early growth, we kept entering and leaving these expansionary and contractionary phases. You have this core idea, but as you’re trying to solve the problem, you’re coming up with three, five, 10 different possible solutions that could all be applicable. So, the expansionary period would be always where we would