Signal › Contributors

Andrei Kozlov

I was a propulsion engineer at SpaceX from 2016 to 2021, working on the Raptor engine program. For those unfamiliar, Raptor is the full-flow staged combustion engine that powers Starship. It's one of the most complex rocket engines ever built, and my team worked on the turbopump assembly, which is the component that feeds propellant into the combustion chamber at the pressures and flow rates required for a full-flow cycle.

Working at SpaceX was the most intense experience of my life. The pace is relentless. The expectations are absurd. The attrition is real: most people burn out within two to three years. I lasted five because I genuinely loved the work. There's something addictive about solving engineering problems where the consequence of failure is a $200M vehicle exploding on the pad. It focuses the mind.

What most people don't understand about SpaceX is that it's not primarily a rocket company. It's a manufacturing company. The engineering breakthrough of Falcon 9 wasn't the reusable first stage (though that's impressive). It was the ability to build rockets at a cost and cadence that made reuse economically rational. SpaceX launches cost roughly $2,700/kg to LEO. The industry average before SpaceX was $54,000/kg. That 20x cost reduction created the commercial space economy.

After SpaceX, I spent a year at NASA's Jet Propulsion Laboratory working on propulsion systems for deep space missions. JPL is the opposite of SpaceX in almost every way: slower, more methodical, more risk-averse, and more focused on scientific objectives than commercial ones. Both cultures produce extraordinary engineering. The trade-offs are different.

I left aerospace in 2022 because I wanted to write about the commercial space industry and deep tech more broadly. The coverage of these sectors is terrible. Space journalism is either "Elon said a thing" or "cool rocket picture." Nobody covers the business economics: what does a satellite constellation's unit economics look like? What's the actual addressable market for in-orbit manufacturing? How do launch costs flow through to downstream applications?

My writing for Signal covers space and deep tech with an emphasis on engineering economics. I write about launch vehicles, satellite constellations, nuclear energy, quantum computing, and fusion, the technologies that take decades to develop and billions to commercialize. The common thread is that these industries require understanding both the physics and the business model, and most analysts only understand one.

I'm based in Munich, Germany. I grew up in Yekaterinburg, Russia, studied at TU Delft in the Netherlands, and moved to the US for SpaceX. I'm a private pilot, I climb in the Alps on weekends, and I build model rockets with my son, which he finds less impressive now that he knows I used to build real ones.

Experience

Articles by Andrei Kozlov (5)

Brett Adcock Just Raised $700M for a 70-Person Startup. The AI Device Race Is Back.Hark raised $700M at $6B with 70 employees. Adcock's startup builds AI models and hardware together from day one. Inside the bet, the Nvidia-AMD inves · May 23, 2026Building the In-House AEO Team: Org Charts, Roles, Budgets, and Reporting LinesThe companies winning AI search have built dedicated AEO functions. Here is what those teams look like, what they cost, and how the highest-performing · May 25, 2026Multi-Touch Attribution in the AI Search Era: Why Last-Click Is DeadOG and Twitter Card metadata now feeds AI summaries on iMessage, Slack, Discord, X, and the major assistants. Most marketing sites leak 30-50% of cita · May 25, 2026ChatGPT GPT Store Submission Strategy: Brand Visibility Inside the ConversationMainland China's AI search ecosystem runs on parallel rails — Baidu Ernie Bot, Tencent Yuanbao, ByteDance Doubao, Moonshot Kimi, DeepSeek, and Zhipu G · May 25, 2026AEO Agency M&A Heat Map: Who's Acquiring, Who's Getting Bought in 2026Survey data across 412 B2B and B2C marketers shows AEO budgets averaging 11.3% of total marketing spend in 2026 — with a 5x spread between ad-hoc and · May 26, 2026