Trajectory raises $45M led by Sequoia to keep retraining open models on feedback from your app

Trajectory announced a $45 million Series A at a $300 million valuation on November 2, led by Sequoia with Nvidia, Bessemer Venture Partners and Bain Capital Ventures, TechCrunch reports. The company is about a year old and has 18 people. Its three co-founders met at Stanford; Ronak Malde was among the researchers Google DeepMind hired from Windsurf, Michael Elabd also worked at DeepMind, and Arjun Karanam worked on generative foundation models at Apple.

The product is continual learning as a service. You start with an open-source model, connect Trajectory to your application, and it collects feedback on how the model does and retrains it on that feedback. The pitch is that a lawyer or analyst defines what good looks like, instead of a company hiring a research team or forward-deployed engineers to fine-tune.

Harvey and Rogo use it to post-train open models on tasks like reviewing one type of contract or valuing a company with a specific method. Decagon, Mercor and Nvidia are also customers. With the round, the post-training platform is publicly available.

TechCrunch names the competition: Thinking Machines Lab and its Tinker fine-tuning platform, Applied Compute (valued at $3.25 billion last month) and Prime Intellect.