Why Physical AI Is the Next Platform Shift
In a Nutshell
Eric Landau's Encord is building the data infrastructure for physical AI, which targets the 80% of economic activity involving physical manipulation and movement in the real world. The company evolved from vision-focused annotation to handling multimodal robotics data across petabyte-scale datasets, serving autonomous vehicles, manufacturing, and logistics applications. Key lessons include persisting through multiple failed sales hires before finding product-market fit organically, maintaining dual London-Bay Area operations to access talent and customers, and treating startup volatility as a manageable roller coaster rather than a source of stress.
These notes were generated by AI and may contain inaccuracies.
Eric Landau is the co-founder of Encord, building the data layer for physical AI. His background includes big data particle physics followed by 10 years as a quant before creating Encord. The common thread connecting these experiences is the trajectory of AI development over his career. When he started in physics doing particle physics, systems were based on the physics itself, filtering data based on particle momentum and movement. In high frequency trading, the approach involved thinking about market factors and crafting strategies before applying machine learning models. The current paradigm is simply throwing data into a large machine learning system.
Eric quit his quant job during COVID to start a company. The month he left, the market went crazy with oil trading negative, extreme volatility, and Wall Street bets activity. His desk made more money that month than in the previous 3 years combined. He quit not for money but because he and his co-founder believed AI represented the technological paradigm shift of their generation - the early days of the internet and computing. After starting the company, his existential questions disappeared and were replaced with concrete problems about providing value to customers.
Encord was created in 2021. The company spent a couple years in the desert before ChatGPT changed the conversation tenor. Product-market fit for Encord came gradually rather than through a single binary threshold moment. The product continuously improved, and after ChatGPT, the market started coming to them faster. Product-market fit was recognized retrospectively when a sale closed in the Gong channel with a company they didn't know, indicating they had achieved organic traction without being intensely involved in every sales process.
Sign in to read the full notes
Get access to AI-generated notes, topic timestamps, and more.