Stay ConnectedFew companies make it from pilot to production in the defense market. Those who do often change the industry in the process.
How do they do it? What lessons can startups take from their trials, successes, and failures? Crossing the Valley tells the stories of the trailblazers who are forging a new path for America’s defense.
Copyright: © Noah Sheinbaum
About Phil Duong
Phil Duong is the CEO and co-founder of Noda AI. The son of Vietnamese immigrants who arrived in the U.S. with little money and speaking little English, he grew up in a family of service members who had fought for their freedom. His heritage left him with a deep sense of patriotism and a drive, as he puts it, to repay the country “multiple lives’ worth.” He won a spot at West Point but walked away from it when a childhood medical issue blocked him from branching aviation, calling the Marine Corps instead because they needed pilots and were willing to give him a shot. After leaving active duty, he built a career at enterprise-AI company C3, rising to general manager of its federal practice before leaving to start Noda. He frames himself first and foremost as “a Marine and patriot,” a self-description that shapes how he runs the company and what he’s willing to say in public.
About Paige Craig
Paige Craig is the founder and general partner of Outlander VC. Growing up homeless, and raised in what he describes as a violent world, he was named for a grandfather who rose from enlisted soldier to an officer under General Patton. Paige was recruited to West Point — where he claims the distinction of being the most-punished cadet in his class — and left in his third year, recognizing he was “the wrong guy” for the institution’s barracks mentality. He moved on to the Marine Corps and the intelligence community, and in 2003 built a private military company operating at and beyond the forward edge in Syria, Afghanistan, and Africa, which he sold. As an investor, his entire lens is the founder and the mission, evaluated against a proprietary 38-point framework built around vision, intelligence, character, and execution.
About Noda AI
Noda AI is a startup building the orchestration and decision layer for autonomous warfare. Rather than building drones or vehicles, Noda aims to overlay government-designed tactics and strategy onto the autonomous platforms other companies are building. Their software seeks to re-create what a strike cell or operations center does today, fast enough to run in real time as robots take over more of the battlefield. The idea is that battle plays for defeating advanced threats need to be baked into deployable algorithms in advance, ready to push down to autonomous boats, planes, and submersibles the moment competition turns to conflict. The company started as a five-person team that prototyped for roughly a year unpaid, and is now working across what Phil will only describe as “very topical” combatant commands. Find them at NodaIntelligence.ai.
About Outlander VC
Outlander VC is an early-stage venture firm focused on writing first checks into exceptional, often low-signal founders others haven’t yet recognized. The firm runs a high-volume, high-selectivity funnel. Paige says that over 10,000 founders come in through the top of the funnel each year, with roughly 5,000 passing AI filters to a first call, a few hundred meeting the team in person, and 10 to 12 ultimately receiving funding. Paige keeps more than half his calendar deliberately unscheduled so he can take long walks with founders and go deep on the person rather than the pitch deck. The firm now has seven partners Craig has trained.
For more about…
* Noda AI: https://www.nodaintelligence.ai/
* Outlander VC: https://outlander.vc/
* Crossing the Valley: www.valleycrossers.com
Follow:
* Phil Duong: LinkedIn
* Noah Sheinbaum: X | LinkedIn
When it comes to the defense tech sales G.O.A.T. debate, John Doyle is one of the most oft-mentioned names in the business. From his days at Palantir, to his robust advisory portfolio, John is both successful and generous.
His appearance on Crossing the Valley has been a long time coming!
This week, John joined us to talk about Cape — a privacy-first cellular carrier he started in 2022 after a decade at Palantir and a tour as a Green Beret communications leader.
The simple version of the story is “ex-Palantir guy builds secure phone company.”
The truth, of course, is a whole lot more interesting.
For more on…
Cape: https://www.cape.co/
John Doyle: https://www.linkedin.com/in/john-doyle-48633227/
Crossing the Valley: LinkedIn | YouTube
How Nominal Turned the Pentagon’s Slowest Gate Into a Growth Engine
Featuring Cameron McCord, CEO and co-founder of Nominal
About Cameron
Cameron McCord has seen the defense ecosystem from more seats than most. He started his career as a young submarine officer, operating, in his words, with “the best of 1980s technology” and learning what it feels like to be beholden to antiquated tech when the mission is on the line. He carried that frustration into operating roles at Anduril and Saildrone, where he watched fast-growing hardware companies scale and saw the same testing problem appear at every one of them. Then he went to Lux Capital, where he saw the same pain across a thousand more companies and turned a hunch into a thesis.
Today, that thesis is a company called Nominal, one of the fastest growing, well-capitalized, and highly touted defense tech companies in the industry.
About Nominal
Nominal is a software and data platform for hardware testing operations. The hardware test market is a bit anathema for us non-technical folks. But the reality is, anyone who builds anything in the real world, whether for a commercial market or the warfighter, has to test and validate it. Until recently, that was mostly done via spreadsheets, MATLAB scripts, and hard-to-reach data. Nominal took many of the processes that befuddle founders and seasoned executives alike, and automated them. The company was incubated at Lux Capital. They intended to serve both commercial and defense customers from day one. And they’ve been on a tear, recently announcing a sole-source $53 million IDIQ for the Air Force Test Center. Today, Nominal is active in the Navy at Pax River, developed partnerships with MIT, and has some exciting news with DARPA.
After speaking time with Cam, I left convinced that Nominal really is going after the single biggest (remaining) roadblock to meaningful acquisition reform. If we don’t solve test, we cannot meaningfully shrink the divide between cool ideas in R&D and capabilities that are validated and safe to deploy. If we’re entering a tech boom cycle (and if you’re even a slightly AI-pilled reader, it’s not hard to imagine that we are), then the need to quantify risk, readiness, and the value of a test (especially one you can skip) becomes a whole lot clearer. Cameron’s whole approach is a bet that the way across the valley of death is not to test less or test faster in a vacuum, but to finally be able to prove, with data, exactly how much testing is enough.
For more on Nominal: LinkedIn | Website | X
For more Crossing the Valley: valleycrossers.com | YouTube
How Machina Labs Compressed 4-Year Lead Times Into 1 Week
About Ed Mehr
Ed Mehr is the CEO and co-founder of Machina Labs. A builder since childhood (he attended a school that emphasized welding, carpentry, and craft) Ed started his career in software before crossing over into hardware as an early engineer at SpaceX in the early 2010s. The cultural shock of moving from “ship a prototype in days” software to “wait months for a single iteration” hardware became the founding wound that drove him toward agile manufacturing. After SpaceX, he joined Relativity Space to work on 3D-printed launch vehicles, then founded Machina Labs to build the missing infrastructure he believed was the real bottleneck to America’s next industrial age: a general-purpose manufacturing platform that didn’t require part-specific tooling or CapEx. He’s a self-described capitalist and car guy, and prints small batches of his own shirts when an idea strikes him. Kinda badass.
About Machina Labs
Machina Labs is building “RoboCraftsman.” These are autonomous robotic cells that form sheet metal parts the way a potter shapes clay. Two robots come at a flat sheet of metal from either side and progressively deform it into complex 3D geometries, eliminating the need for the dies, molds, and multi-story stamping presses that have defined sheet metal manufacturing for a century. The company serves a dual-use customer base spanning commercial (Toyota is both a major customer and an investor) and defense (Air Force depots, missile primes, NASA). Machina has already deployed production cells at Air Force depots, where they’re making C-130 panels and other legacy aircraft parts that previously required hunting down donor airframes in junkyards or waiting four years for new tooling. The company is roughly 70–80 people, hiring aggressively, and approaching the end of qualification testing with parts on track to fly soon. Machina is now entering its scale phase, and Ed teases that a missile-program prime announcement is imminent…
For more on Machina:
* Careers: machinalabs.ai/careers
For more Crossing the Valley:
* Follow Noah: https://www.linkedin.com/in/noahsheinbaum/
* Substack: valleycrossers.com
* YouTube: https://youtu.be/_WN-GJaCbGs
Rebuilding America’s Thermal Imaging Supply Chain
About John Hong and Tallis Chang
John Hong is CEO and Tallis Chang is COO of Obsidian Sensors. The two have worked together for over 40 years, with backgrounds in aerospace before joining Qualcomm. At Qualcomm, they were part of the team that spent nearly a decade developing low-power displays using MEMS (Micro Electro Mechanical Systems) built on glass substrates. When Qualcomm decided the program no longer fit its core business, John and Tallis were given the runway to find a new application for the manufacturing know-how they had built. That pivot became Obsidian Sensors.
About Obsidian Sensors
Obsidian Sensors, headquartered in San Diego, manufactures microbolometers — the core sensor inside thermal imaging cameras — using a fundamentally different platform than the rest of the industry. Where incumbents builds them on silicon wafers, Obsidian builds these sensors on display glass, leveraging LCD fabrication infrastructure in Japan and Taiwan that already produces millions of phone and watch displays. This shift dramatically reduced unit cost while preserving sensor performance.
The company spun out of Qualcomm in 2017 with seed capital, transferred equipment, seven team members, and five patent applications. It initially targeted the automotive market before pivoting to defense after receiving a DIU OTA at the start of COVID. Today, Obsidian is a critical supplier of thermal cameras to the US drone ecosystem, with production scaling from 5,000 units in 2025 to a planned 1 million units per year by the end of 2026 at its Top Gun Street facility in San Diego. A Phase 2 facility targeting 10–20 million units annually is planned for 2028.
Follow John: https://www.linkedin.com/in/john-h-973b372/
Follow Tallis: https://www.linkedin.com/in/tallischang/
Learn more about Obsidian: https://www.obsidiansensors.com/
Follow Noah: https://www.linkedin.com/in/noahsheinbaum/
Follow Crossing the Valley: YouTube | LinkedIn
About Brian Spurlock
Brian Spurlock is VP of Growth and Strategy at Persistent Systems, and a self-described “boomerang” employee, meaning he’s had two separate tenures at the company. He came up as a Special Operations acquisitions officer at Fort Bragg, which means he’s seen the Persistent Systems story from both sides: as the government buyer evaluating the technology and as the company executive selling it. He left for a stint at AWS, returned, and is now the one figuring out how to bring tactical mesh networking from Special Operations into Army divisions. He is based in the Fort Bragg / Pinehurst area of North Carolina.
About Persistent Systems
Persistent Systems builds tactical mesh networking hardware and software: specifically, the Wave Relay and Cloud Relay systems, delivered on hardware called the MPU (currently MPU5, with MPU6 in development). The idea is simple and hard: create a mobile RF network that moves with operators, passes real data at real speeds, and requires zero thought from the person holding the radio. Founded in 2007 by two Johns Hopkins PhDs — Herb Rubens and Dave Holmer — the company has never taken outside capital. It manufactures in Manhattan. It has customers across U.S. Special Operations, conventional Army (currently scaling into division-level exercises with 4th Infantry Division), the Air Force, NATO allies, and public safety. It is on the verge of Army program-of-record status after nearly 20 years.
Follow Brian: https://www.linkedin.com/in/spurs/
For more Persistent Systems: https://persistentsystems.com/
For more Crossing the Valley:
* LinkedIn: https://www.linkedin.com/company/crossing-the-valley/
* YouTube: https://www.youtube.com/@crossingthevalley
Building the Physical AI Platform for the Industrial Base
ABOUT ARIYAN KABIR
Ariyan Kabir is the co-founder and CEO of GrayMatter Robotics. He started the company in early 2020 alongside two co-founders, building on PhD research conducted at USC’s Center for Advanced Manufacturing, where he and one co-founder studied the intersection of robotics and AI. The third co-founder, chief scientist Satyandra K. Gupta, has spent more than 35 years working at the intersection of AI, robotics, and manufacturing and sits on multiple national boards.
Ariyan’s framing of the problem is shaped by a realization that came from the academic environment: that manufacturing (long assumed to be “solved”) is in fact one of the largest unaddressed application areas for modern AI. The US is currently short half a million skilled manufacturing workers, a gap projected to reach four million in seven years and put $2 trillion of annual GDP at risk.
ABOUT GRAYMATTER ROBOTICS
GrayMatter Robotics builds the autonomy layer for high-mix, high-variability manufacturing, the roughly 90% of factory work that cannot be addressed by traditional pre-programmed robotics. The company’s robots perform tool-manipulation tasks: sanding, grinding, polishing, buffing, coating, blasting, and inspection. Underneath the robots sits the company’s real differentiator: a foundation model for materials and processes, trained on what Ariyan describes as the largest manufacturing process dataset in the world: 7 petabytes of multimodal sensor data spanning 14–18 modalities per cell.
The business is now six years old. This interview was recorded at the company’s 100,000 sq ft Physical AI Experience Center in Carson, California, which is the company’s fifth facility. Today, GrayMatter’s customers include Boeing, Raytheon, Oshkosh, Caterpillar, Riddell, the US Navy, the US Air Force, and Huntington Ingalls Industries, where GrayMatter is a partner on the HYPR (High Yield Production Robotics) program designing the next-generation American shipyard.
The company sells outcomes, not hardware. Robots are deployed on a subscription basis with multi-year commitments, and performance compounds quarterly through software updates, recipe upgrades, and hardware refreshes.
KEY TAKEAWAYS
1. Sequence the market. Commercial first, defense second.
GrayMatter spent its first three-plus years building almost entirely in the commercial world, despite the obvious dual-use relevance of factory automation. Their reasoning was that defense customers tolerate zero failure, and an unproven autonomy stack cannot debug itself inside an aircraft program. By the time the company turned toward national security work, they had already validated the technology, the business model, and the deployment muscle.
2. The business model can be the product.
Performance Composites — GrayMatter’s first customer — had been searching for a robotic sanding solution for five years before the conversation with GrayMatter. IT wasn’t so much a capability they were missing as it was a different risk appetite: manufacturers cannot tie up millions of dollars in Capex on unproven technology. GrayMatter’s response was to invert the deal entirely: zero upfront cost, deployment first, subscription payments only after the system was producing value in production. In exchange, the customer committed to a multi-year contract. The structure became the wedge that opened the entire Fortune 100.
3. Adoption became the geostrategic moat.
Asked how 7 petabytes of US manufacturing data compares to South Korea or Shenzhen, Ariyan reframed the question. The race isn’t about dataset size. The US ranks near the bottom of the top 10 in robot-to-worker adoption ratios; South Korea is #1 and China is climbing fast. Tonnage-wise, China has built more ships in the last 12 months than the US has built since WWII. This reframing has direct strategic implications: a defense industrial base strategy that focuses on technology sovereignty without obsessing over diffusion velocity will lose to one that does the opposite.
4. GrayMatter builds AI into every layer.
GrayMatter identifies eight distinct layers of engineering work that sit between a customer problem and a delivered outcome: industrial engineering, solution engineering, process recipe engineering, cell design, tool/fixture design, implementation, sustainment, and digital-twin operations. Most robotics companies put AI in one layer: the robot. This means the other seven can become bottlenecks. But GrayMatter is now deploying domain agents at every layer, coordinated by an orchestration layer the company calls Factory Super Intelligence (FSI). The focus on under-appreciated elements of the solution set has helped them differentiate.
5. Hire from the intersection. The mixing room doesn’t work.
GrayMatter’s most expensive hiring lesson was that you cannot assemble a great physical AI team by recruiting the best AI specialists, the best manufacturing engineers, and the best roboticists into a room. The translation cost is too high; each discipline optimizes for its own definition of the problem. What does work is hiring people who already live at the intersection of at least two of the three domains, then leveling them up on the third. The company’s head of customers came from Tesla. Their head of aerospace and defense was two steps below Boeing’s CTO. Their head of hardware also came from Tesla. This was a counterintuitive lesson, but one that stuck with me most from the conversation.
Follow Ariyan: https://www.linkedin.com/in/ariyankabir/
For more GrayMatter: https://factory.graymatter-robotics.com/
For more Crossing the Valley: www.valleycrossers.com
About Ben
Ben Cipperley spent 26 years in the Navy. The first 18 he served as an Explosive Ordnance Disposal office, with three tours across the Indo-Pacific, three years between Iraq and Afghanistan, and command in Guam. Which is to say, he has more experience than most.
What makes his background unusual isn’t the operational experience; plenty of veterans make the transition. It’s the Pentagon chapter that came after. A fellowship at the Stimson Center led to a resource sponsor role writing the Navy’s budget, which led to being pulled onto the team drafting the 2022 National Defense Strategy (mid-process, while Russian forces were marshaling in Belarus). Which led to working under the CNO on what became the Navy’s first force design document in 50 years. It’s a classified vision for the fleet in 2045 that, he notes, had quite a few robots in it.
He put an “Open to Work” banner on LinkedIn when he retired. The CEO of Havoc called him 20 minutes later.
About Havoc AI
Havoc AI is a collaborative autonomy software company. The pitch is deceptively simple: one operator, thousands of autonomous systems. The technology is the connective tissue, the software layer that lets a surface drone, an aerial drone, and a ground vehicle share a common operational picture and respond to commands from a single interface, rather than requiring separate operators, separate control systems, and separate data pipes for each platform.
The company was born from a conference convened by then-Undersecretary of Defense Heidi Shyu around the concept of “Hellscape,” IndoPaCom’s deterrence concept for a Taiwan scenario involving massive swarms of autonomous systems. The question in the room wasn’t whether the platforms existed. It was: how do you make thousands of them work together when a human brain can only track so many things at once? Founders Paul Lwin and Joe Turner quit their jobs the next day and started building the answer.
Early validation came through Silent Swarm, an exercise run by Naval Surface Warfare Center Crane. Havoc showed up with 12 prototypes of their Rampage platform. Representatives from NIWC Atlantic watched, then bought the prototypes on the spot. Those units shipped to Portugal to support Task Force 66. That’s the loop Havoc is now trying to close, and why they recently acquired Mavrik (air) and Teleo (ground) to make good on the multi-domain part of the pitch.
The business model is worth understanding clearly. The government is bad at buying software. It tends to need software wrapped in a hardware container before it can conceptually acquire it. Havoc has internalized this and partnered with eight different maritime platform builders, plus air and ground hardware partners, so that their autonomy stack arrives pre-integrated with best-of-breed hardware. They’re not trying to be the OS layer that makes everyone else’s platforms work together. But they’ll sell you the hardware too, because that’s how you get the check.
For more about Havoc AI: https://havocai.com/
Follow Ben: https://www.linkedin.com/in/bencipperley/
For more Crossing the Valley: valleycrossers.com
Follow Noah: https://www.linkedin.com/in/noahsheinbaum/
Crossing the Valley explores the journey from proof of concept to production in defense technology. New episodes weekly.
About Luke Fischer
Luke Fischer is the CEO and co-founder of SkyFi. He is a former U.S. Army helicopter pilot and Special Operations veteran who transitioned through DIU (Defense Innovation Unit), Uber Elevate (as Head of Flight Operations), and Joby Aviation (leading government operations) before founding SkyFi in late 2021. He also spent time as an Entrepreneur in Residence in VC, which gave him direct exposure to what gets funded — and what doesn’t. That multi-domain journey across military, Big Tech, and venture is core to how he built SkyFi.
About SkyFi
SkyFi is a geospatial intelligence marketplace and platform that aggregates satellite imagery and analytics from hundreds of commercial satellite operators — making it accessible to any customer, anywhere, via web or mobile app. Rather than owning or launching satellites, SkyFi sits at the aggregation layer: abstracting away which provider owns the satellite and letting users task imagery or pull analytics with a button press. The business is currently approximately 80% government / 20% commercial, serving customers ranging from top global hedge funds and agriculture companies to Special Operations Forces, NATO, and allied governments across Asia. SkyFi closed its Series A in December 2024.
For more SkyFi: skyfi.com
For more Crossing the Valley: valleycrossers.com
Follow Luke: https://www.linkedin.com/in/lukemfischer/
Follow Noah: https://www.linkedin.com/in/noahsheinbaum/
About Todd Stiefler
Todd Stiefler is VP of Enterprise at WHOOP, overseeing the company’s B2B and B2G business lines. He brings an unusually cross-functional background to the role: a decade as a Senate defense staffer focused on missile defense and strategic nuclear deterrent, followed by product and commercial leadership roles at GE’s industrial IoT division, LMI, and Palantir. A self-described “tech nerd” and ultra-marathoner who finished the Leadville 100 in 27 hours, Todd found WHOOP the way most people find things that change their lives — by accident, scrolling LinkedIn late one night. He’s been building the company’s defense business ever since.
About WHOOP
WHOOP is a Boston-based wearable technology company focused exclusively on human performance and health optimization. Unlike general-purpose smartwatches, WHOOP does one thing: continuously measures high-fidelity physiological data — heart rate variability, resting heart rate, respiratory rate, skin temperature, and strain — and translates it into actionable metrics around sleep, recovery, and readiness. Its early customers included elite professional athletes, and it has since grown into a mass-market product worn by hundreds of thousands of consumers and, increasingly, military personnel. WHOOP is actively pursuing the DOD enterprise market, partnering with programs like SOCOM’s POTD and the Army’s Holistic Health and Fitness (H2F) initiative. The company’s enterprise division, led by Todd, handles everything from individual unit programs to large-scale institutional deployments.
Key Takeaways
1. Build for your most demanding user first.
In contrast to the “minimally viable product” framework, WHOOP’s early decision to build exclusively for elite athletes — with LeBron James and Michael Phelps among their first 100 customers — wasn’t a marketing stunt, but a product strategy. By targeting the most demanding performance environment, they built something rigorous enough to eventually serve everyone else. That same logic attracted special operators before conventional forces, and conventional forces before institutional DOD programs. They solved for the “edge cases” first rather than the average user, because they were solving their distribution challenge through their product design.
2. But the right product alone isn’t enough… you still have to solve the process.
Todd was direct about what WHOOP got wrong in its early defense push: they had a best-in-class product but hadn’t built the surrounding infrastructure. No contract vehicles. No data sharing agreements. No pricing architecture tuned to government procurement. “The DOD buys from people that solve their process, not their problem.” Getting that right — not just shipping a great device — is what actually enables scale. So while many of us laugh at high-ceiling IDIQs without real funding, the vehicle is a key ingredient.
3. The wearable itself is actually not the intervention.
This was one of the most important points from the episode: motivated, self-selected consumers who buy WHOOP on their own are signaling their desire to change their behavior. Warfighters handed a device in a conference room, with no context, no coaching, and no program design, often won’t. Even though the average warfighter may be more attuned to their body than the average civilian, in this case, if the warfighter doesn’t opt in, the product won’t deliver. This insight is why Todd never sees WHOOP going to a “push to all” rather than a “pull” based sales motion. If it’s foisted upon people, it will not produce the desired effect.
4. In federal sales, relationships and strategic partnerships beat conference booths.
Todd assesses that in the defense market, niche events where you can have real conversations, and partnerships with athlete management system providers who can pull WHOOP into their stack, have been the most valuable sources of leads. Trade show booths? Much less effective. As always, your mileage may vary, but it’s another reinforcing data point suggesting deep, niche relationships beat a larger breadth with high impressions.
5. Sell sustainable programs, not hardware deployments.
Todd’s north star for WHOOP’s next 12 months is helping customers build programs that will still be running in five years. That means bringing in WHOOP’s human performance experts and solution architects early, co-designing integration with existing data platforms, and being the vendor who asks, “What are you actually trying to achieve?” before taking the purchase order.
For more Crossing the Valley: valleycrossers.com | LinkedIn
For more on WHOOP: WHOOP.com
Follow Todd: https://www.linkedin.com/in/toddstiefler/
Follow Noah: https://www.linkedin.com/in/noahsheinbaum/
Our TOPPODCAST Picks 
