Trajectory Raises $15M for Continual Learning Platform

Trajectory, a research lab and product company pioneering continual learning, has raised $15 million to build a platform that continuously trains large-scale agentic models using real-time product usage signals, shifting the industry from static training to a living learning system. Backed by investors including Conviction, Bessemer Venture Partners, Radical Ventures, Jeff Dean, and Fei-Fei Li, the startup aims to reduce model drift and enable products that grow smarter over time.

The founding team, drawn from DeepMind, OpenAI, Apple, and Meta Superintelligence, brings deep expertise in scaling AI systems. Trajectory’s clients already include Clay, Harvey, Decagon, Mercor, and Rogo, with some deployments live in production environments. By capturing signals from user interactions—such as corrections, retries, and edits—the platform continuously refines large-scale agentic models, turning AI systems into adaptive, self-improving tools.