Why Physical AI Is the Next Platform Shift

Landau connects his work in particle physics, high-frequency trading, and AI through the shift toward scaling machine learning with more data and compute. He left a lucrative quant job during the volatile early days of COVID because he and his co-founder believed AI was a major technological shift and wanted to build something useful.

Encord spent a couple of years struggling to persuade customers that AI mattered. ChatGPT made people more receptive, while the product continued improving gradually. Landau says there was no single moment of product-market fit; one sign came when a sale closed without the founders knowing what the customer did.

Encord began with computer vision, then expanded to multimodal data as customer needs and the market shifted toward physical AI. Landau points to robotics, autonomous vehicles, logistics, and manufacturing, and estimates that 80 percent of economic activity involves moving or manipulating physical things. The company tracks emerging needs by talking with customers, prospects, and other founders.

Autonomous driving is closer to mass production, while robotics is a growing area across humanoid, consumer, and manufacturing applications. Encord collects data in a Bay Area facility and ingests customer production data, including video, sensor data, audio, and language. Its product helps select, manage, annotate, enrich, and evaluate data; Landau says the company serves hundreds of customers and handles multiple petabytes.

Encord built operations near its many Bay Area customers while Landau remained in London, which he sees as a strong source of AI and engineering talent. The company learned sales hiring through repeated attempts and became quicker to act when a hire was not a fit. Landau advises founders to accept the emotional swings of startup life and learn to influence their own reactions.