Trajectory raises $15M in seed funding for continual learning platform
What's the deal? Trajectory, a San Francisco-based research and product lab, has raised $15M in seed funding to build a platform for continual learning in AI. The round was led by ConvictionDealroom has a profile for this one. Try Dealroom →, with participation from Bessemer Venture Partners, Radical Ventures, BoxGroup, and angel investors including Google's Jeff Dean and Stanford's Fei-Fei Li. Founders of Notion, Dropbox, Hugging Face, and Braintrust also joined.
Led by chief executive officer Ronak Malde, the company is building what it calls the first platform for continual learning — an intelligence layer that replaces static AI models with systems that improve with every user interaction. It already partners with AI-native companies such as Clay, Harvey, Decagon, Mercor, and Rogo.
The funds will go toward expanding operations and development.
Why now? Most AI models today are frozen at training time. They don't learn from the data they process once deployed, which limits their usefulness for companies whose products depend on improving over time. As AI-native startups proliferate, the gap between static models and the dynamic intelligence these businesses need is becoming a pressing problem.
What could go wrong? Continual learning is a notoriously hard technical challenge. Models that update on the fly risk catastrophic forgetting — losing previously learned knowledge as new data comes in. Trajectory will also need to convince customers that a third-party intelligence layer is worth the integration cost and potential vendor lock-in.
The investor list is impressive, but $15M is modest for a company tackling a research-heavy problem with infrastructure ambitions. Scaling both the science and the go-to-market simultaneously will test the team's focus.
The signal: The round reflects growing investor appetite for the picks-and-shovels layer beneath AI applications. Rather than building another model or another chatbot, Trajectory is betting that the real value lies in making deployed AI systems continuously smarter. The backing from prominent AI researchers like Dean and Li suggests the technical thesis has credibility. If continual learning works at scale, it could become foundational infrastructure for the next generation of AI-native software.
Read more: FinSMEs
