Trajectory raises $45 million at a $300 million valuation

Trajectory has raised $45 million in a Series A round at a $300 million valuation. The one-year-old startup said the round was led by Sequoia, with participation from Nvidia, Bessemer Venture Partners and Bain Capital Ventures.

Founded by three Stanford classmates, Trajectory aims to help businesses train AI models for specialized work without relying on in-house research teams or expensive consultants. Its platform connects to a company’s applications, collects feedback on model performance and uses it to retrain models over time.

Co-founder Ronak Malde said specialists such as lawyers and doctors can use the platform to improve models for their own tasks. Companies including Harvey and Rogo use it to post-train open-source models on work such as reviewing particular kinds of legal contracts or valuing companies with specific methods. Other customers include Decagon, Mercor and Nvidia. Trajectory is now making its platform publicly available.

The approach has similarities to recursive self-improvement, but co-founder Michael Elabd said keeping people involved in training can help avoid unintended outcomes from autonomous improvement. Trajectory competes with Thinking Machines Lab, Applied Compute and Prime Intellect. The 18-person startup’s other co-founders are Elabd and Arjun Karanam. Malde and Elabd previously worked at Google DeepMind, and Karanam worked on generative foundation models at Apple.