Startup Trajectory Raises $45 Million to Automate AI Model Fine-Tuning

The Round and the Founders

Trajectory, a startup founded about a year ago by three Stanford classmates with experience at Google DeepMind and Apple, has raised $45 million in a Series A round at a $300 million valuation. The round was led by Sequoia, with participation from Nvidia, Bessemer Venture Partners and Bain Capital Ventures. This was reported by TechCrunch. The 18-person company employs alumni of OpenAI, Stripe and Netflix.

What the Product Does

The problem the startup solves: modern AI models do not learn from their mistakes without additional training, and adapting them to specialized fields such as law and finance usually requires in-house research teams or expensive consultants — so-called forward-deployed engineers. Trajectory offers a continual learning approach: a company connects the platform to its applications, the platform collects feedback on the model's performance and uses it for retraining, improving the model on specialized tasks. Co-founder Ronak Malde compared base models to a very smart PhD student who starts a new job every time: 'But the economy needs someone who has been doing that job for 30 years.'

Customers and Public Launch

According to Malde, lawyers, doctors and other specialists will be able to improve models almost automatically themselves, without technical training. The platform is already used by Harvey and Rogo to fine-tune open-source models — for example, on analyzing specific types of legal contracts or valuing companies under a particular methodology. Other customers include Decagon, Mercor and Nvidia. The startup is now opening its post-training platform to the general public. Co-founder Michael Elabd notes that, unlike the idea of recursive self-improvement (RSI), keeping humans in the training loop makes the approach safer.

Competitors

Trajectory is not the only player in this market. The company names as its direct competitors Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, which created the fine-tuning platform Tinker, as well as Applied Compute, valued at $3.25 billion last month, and Prime Intellect, which helps companies build and train their own AI agents. Malde believes Trajectory's main advantage is its team: he himself was among the researchers Google DeepMind acqui-hired from Windsurf, Elabd also worked at DeepMind, and the third co-founder, Arjun Karanam, worked on generative foundation models at Apple. Karanam adds that the startup has developed an interface that makes post-training accessible to non-technical customers.