Trajectory Raises $45M From Sequoia to Let Lawyers and Analysts Retrain Open Models Themselves
Trajectory, an 18-person startup founded about a year ago by alumni of Google DeepMind and Apple, has raised a $45 million Series A at a $300 million valuation, led by Sequoia with Nvidia, Bessemer and Bain Capital Ventures, TechCrunch reported on November 2. Its platform collects feedback from apps built on open-source models and uses it to keep post-training them. Harvey, Rogo, Decagon, Mercor and Nvidia are named as customers.
Trajectory's bet is that specialist companies will want to keep training their own models rather than wait for the big labs to get better at their field. The startup has raised $45 million in a Series A round at a $300 million valuation, TechCrunch reported on Monday, November 2. Sequoia led the round, with Nvidia, Bessemer Venture Partners and Bain Capital Ventures taking part.
What the platform does#
Trajectory sells what it calls continual learning. According to TechCrunch, a business starts with an open-source model and connects Trajectory's platform to its application, which collects feedback on how the model performs and uses it to retrain the model on specialised tasks over time. The pitch is that domain experts, not in-house research teams or forward-deployed engineers, decide "what good looks like for their organization," co-founder Ronak Malde told the outlet.
TechCrunch says Harvey and Rogo already use the platform to post-train open-source models for legal and financial work, such as reviewing a type of contract or valuing a company with a set method. Other named customers are Decagon, Mercor and Nvidia, which is also an investor. Trajectory is now making the platform publicly available.
Who is behind it, and who it competes with#
The 18-person company was founded about a year ago by three Stanford classmates. 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. Elabd argued that keeping humans in the training loop makes the approach safer than fully autonomous self-improvement.
Trajectory counts Thinking Machines Lab, maker of the fine-tuning platform Tinker, which also ships fine-tuning for its open-weight Inkling model, as a direct competitor, along with Applied Compute, which TechCrunch says was valued at $3.25 billion last month, and Prime Intellect.
Limits#
- The round, valuation and customer list come from TechCrunch's interview with the company; we found no investor announcement or second independent report.
- Trajectory has not published results showing how much its retraining improves models for customers.