Trajectory raises 45 million Series A to let non-engineers retrain AI models

Post-training a model used to need a research team. Trajectory wants a lawyer, a doctor or a financial analyst to do it instead. The one-year-old startup closed a 45 million dollar Series A led by Sequoia. The money will go into opening the work of improving models to non-engineers.

TechCrunch reports the round values Trajectory at 300 million dollars. Nvidia, Bessemer Venture Partners and Bain Capital Ventures also invested. The three co-founders are Stanford classmates: Ronak Malde and Michael Elabd came from Google DeepMind, and Arjun Karanam came from Apple. The team is 18 people.

What the product does

Trajectory connects to a company's own application. It collects feedback on how the model behaved. It uses that feedback to retrain the model for a specialist task. The customer tells the platform "what good looks like for their organization," co-founder Ronak Malde told TechCrunch.

The platform targets fields where the specialist and the engineer are two different people. Lawyers, doctors and financial analysts are the examples the company gives. The product sits between the user and the model. It runs the post-training step as a service, so the customer never writes training code.

Named customers

TechCrunch reports Harvey, Rogo, Decagon, Mercor and Nvidia are already on the platform. Harvey is a legal AI tool. Rogo and Decagon work in finance and customer support. Mercor hires contract AI workers. Nvidia uses the platform on its own models.

That list matters because post-training is usually sold as a research service for one team at a time. Trajectory is packaging the same step as a product that any team can turn on. Each of these customers runs a real AI product in production. They now have a way to improve the model without hiring their own training team.

The round and the stage

The 45 million dollars is a Series A. The valuation is 300 million dollars. The company is about a year old. Sequoia led. Nvidia, Bessemer and Bain Capital Ventures took part. The lead investor's name and the valuation together say the round is a priced bet on the product.

TechCrunch's author Marina Temkin frames the pitch as making AI model improvement reach subject-matter experts. The company has not published a customer count or a revenue number. An 18-person team at a 300 million dollar valuation says the round is sized for building and selling the product.

Source

TechCrunch on Trajectory's Series A and the named customers.