Flow Engineering

We're investing in Flow Engineering's Series B because we believe AI can dramatically compress the iteration cycles in hardware design, just like we've seen happen in software development. Flow has the right combination of technical ambition and real-world adoption. Since their Series A last year, Flow has added General Motors, Rivian, Volkswagen Group, and others as customers. At Rivian alone, adoption grew from 40 to 1,500 users in seven months. That kind of expansion shows Flow can become part of the core engineering workflows inside some of the world's largest companies. Once that happens at scale, it'll change what companies can afford to build, how quickly they can test it, and ultimately how quickly physical products get made. We’re pleased to back Pari Singh and the Flow team, alongside many other great investors!

We just raised our Series B. We get to work alongside engineers developing some of the most remarkable hardware products in the world: humanoid robots, hypersonic aircrafts, and satellite constellations. We help those teams iterate faster and bring what they’re building into the world sooner.  Hardware is entering a golden era, with AI at the center. Flow is where engineers and agents build what comes next.  Our ambition is to take hardware development cycles from months to days. If you want to help us do that, we're hiring across engineering, sales, and FDE.

Exciting systems engineering career opportunity in the defense industry (in-person in D.C.) to work on some of the hardest technical problems in national security. This company is looking for someone with deep experience in software, hardware, or FPGA engineering who can apply that expertise at the systems level and take ownership of the architecture. You’ll own the system model as the authoritative source of design intent. Active U.S. Secret security clearance required. DM us if you’re interested, and we’ll share the details.

Mechanical, electrical, structural, manufacturing, systems, firmware, software, autonomy. Every Mytra robot is all 8 at once. Mytra is building material flow automation, making the software and hardware that powers industrial facilities. Instead of the fixed racks, conveyors, and aisles a traditional material flow system depends on, its robots move material in any direction through a dense vertical structure, all driven by software. They lift 3,000 pounds and run at 99.999% uptime. It's some of the most complex hardware in robotics, and a small team is building it. Those layers are all wired together. Change a wheel's material, and it ripples through motor torque, power draw, ESD, and drive accuracy, eventually surfacing as changes to uptime, throughput, and reliability. One change becomes a dozen new requirements somewhere else. Mytra manages all of it on Flow. Requirements, risks, and tests stay in one place, linked, so the team sees what a change impacts before it becomes a problem. Of everything they tried, Flow was the only platform that could keep up with how fast they build. Thanks to Chris Walti, Allison Miller, Arash Narimani, Sasha Rudolf, and Sam Reynolds for being such great partners. Full story linked in the comments.

Our founder, Pari Singh, will be co-hosting a dinner with our friends at Founders You Should Know in SF on July 21 as part of their Summer Dinner Series. We're excited for a candid conversation on hardware, AI, and what the next generation of engineering organizations looks like. Request to join via link in comments.

The AI Systems Engineering Handbook Volume 1 is out. Seven chapters on building AI into your hardware program: I. Introduction to AI Systems Engineering II. Five Levels of AI Systems Engineering III. Introduction to AI Skills and Agents IV. Introduction to Autonomy V. AI Harnesses VI. Low-Hanging Fruit vs. Endgame VII. Safety and Risks Comment "handbook" for the PDF.

100 Million+ Requirements are already managed in Flow across thousands of engineers building the next generation of aerospace, defense, robotics, energy, and autonomous systems. Over the last year, we've worked alongside the world's most ambitious hardware teams to build a new way of engineering. Agents have reached hardware. Flow Agents continuously connect requirements, design, CAD, simulation, code, and testing, helping teams surface conflicts early, propagate changes automatically, and accelerate iteration without lowering the bar. We're excited to share what we've been building. Learn more in the link in comments.

Nitin A. sold his last company to Rippling. We're thrilled to welcome him as Flow's Head of Engineering. He's the most AI-forward engineering leader I've ever met: he’s thinking about how teams work with AI in a way that's five years ahead of most companies. Nitin led data infrastructure at Rippling after they acquired RunX, the company he founded. While he was there, he built an internal console that deploys groups of AI agents to handle data work, cutting tasks that used to take months of repetitive engineering down to hours. Before RunX, he spent a decade as an early engineer at Google, Lyft, and Stripe building systems that served hundreds of millions of QPS. We're doing things today at Flow that nobody else in the industry is attempting yet. Nitin is already reshaping how our engineers build with AI from the inside. We're hiring across full stack, AI, frontend, and infra. DM me or Nitin if you want to build something nobody has built before. 

The secret behind iterative hardware teams bottoms-up speed + top-down rigor. Here are 3 practices that the best teams (SpaceX, Rivian, Joby) that make it work: 1. Integrate is continuous, not on one off phases Integration happens on every change instead of at the end of the program. Every CAD release is checked against the assembly model. Every harness change is checked against the connector mating list. Every requirement change is checked against the verification plan. Interface mismatches surface within hours of being introduced, when one engineer has changed one thing. The result is that integration stops being a high-risk milestone where everything is exercised for the first time. It becomes a continuous property of how the team works. Conflicts get resolved in a conversation, not a six-month redesign loop. 2. Interfaces are continuously negotiated peer-to-peer, not dictated from above The engineers on either side of an interface own the formal specification between them. The mechanical and electrical engineers co-author the harness interface, including connector pinouts, signal levels, and current ratings. The avionics and propulsion engineers co-author the valve command interface, including timing, telemetry rates, and fault modes. This works because the people writing the spec are the people who have to live with it. They know what each side can give and needs, so the specification reflects engineering reality. The systems function tracks the interface registry, but the contract itself is owned by the people closest to the physics. 3. Domain engineers write/own requirements, not just implementation Requirements authorship lives in the domain. The engineer responsible for the propulsion subsystem writes the propulsion requirements, including the performance envelope, the interface budgets, and the verification criteria they will sign against. Cross-subsystem requirements like mass, power, and thermal get negotiated directly with the engineers on the other side. The systems function still exists, but it has shifted from authoring requirements -> maintaining the structure: the requirements graph, traceability links, and verification status across the program. Every requirement has an author who has to build to it, and a systems team that ensures the whole tree stays coherent. 4. AI as the Systems Engineering Glue AI is changes the equation on top down + bottoms up. An AI system reads every requirement, CAD model, simulation, and line of code in a program, builds a dependency graph, and propagates the consequences of any change within seconds. The cross-organizational tracking that used to require an army of systems engineers now happens on every commit. The next decade of hardware will be built by teams that combine bottoms-up speed with top-down rigor.