Trajectory raises $45 million from Sequoia to let experts retrain AI models

Trajectory, a US artificial intelligence start-up founded about a year ago by three Stanford University classmates, has raised $45 million in a Series A funding round that values it at $300 million, it was reported on Monday, 2 November 2026. The round was led by the venture firm Sequoia, with the chipmaker Nvidia, Bessemer Venture Partners and Bain Capital Ventures also investing. Trajectory makes software that, according to the company, lets specialists such as lawyers and doctors improve an AI model by giving feedback on its answers, without needing to write code or run the training themselves. The 18-person company is now making its platform publicly available after early use by a small group of customers.

Most companies that use AI start with a general model built by a big lab. Those models know a little about everything but are not experts in a specific job, such as reviewing a contract or reading a medical report. To make a model better at a narrow task, engineers usually "post-train" it: they collect examples of good and bad answers and use them to adjust the model. This step needs machine-learning specialists, data pipelines and computing power, which many businesses do not have.

Trajectory's pitch is to hand that work to the people who know the subject. A lawyer can mark which answers are right or wrong, and the system turns that feedback into training data and updates the model, almost automatically, according to the company. If it works as described, it could shorten the gap between a company's experts and the AI tools they use. That is a company claim; how well it performs across different fields will become clearer as more customers use it.

The choice of open-source models matters too. Open-source models can be downloaded and modified by anyone, unlike closed models offered only through a company's service. Retraining an open model on a customer's own data lets that customer own the result and run it where it wants, which appeals to firms with strict privacy needs, such as law and healthcare.

For investors, the round reflects how much interest there is in the tools that sit around AI models, not just the models themselves. Nvidia appears in the round both as an investor and, according to reports, as a customer.

For Indian start-ups and engineers, the round is a marker of what investors are paying for: tools that help businesses adapt open models to their own fields, rather than building new models from scratch.

Peepals calculation: A $45 million round at a $300 million valuation means the new investors put in a sum equal to 15% of that valuation (45 ÷ 300 = 0.15). Reports do not say whether the valuation is measured before or after the new money; if it is after, the new investors own about 15% of the company. For an 18-person team, the valuation works out to about $16.7 million per employee (300 ÷ 18 ≈ 16.7). Our calculation from the reported figures.

Short-term: Trajectory will use the funding to grow its team and open its platform to more customers.

Long-term: If tools like this work well, companies may rely less on in-house machine-learning teams to adapt AI models, and more on their own subject experts.

Who is affected: Businesses deploying AI in specialised fields such as law, finance and healthcare; machine-learning engineers; and developers of open-source models.

Questions and Answers