Trajectory, a company launched roughly a year back by three former Stanford classmates, has revealed a $45 million Series A round that values it at $300 million, per [TechCrunch](https://techcrunch. com/2026/11/02/trajectory-raises-45m-to-make-ai-model-training-accessible-to-nontechnical-experts/). Sequoia led the round. Nvidia, Bessemer Venture Partners and Bain Capital Ventures also took part. The company has 18 employees.
What the product does
The problem Trajectory targets is that today's models do not learn from their own mistakes unless they are trained again. Companies that want a model to handle a specialized field such as law or finance usually fine-tune a general model. That tends to require an internal research team or paid outside engineers.
Trajectory wants to automate that loop. A business starts with an open-source model and connects the platform to its applications. The platform gathers feedback on how the model performs and uses it to retrain the model, so it improves at specialized tasks over time. The company calls the approach continual learning.
Co-founder Ronak Malde said lawyers, doctors and other specialists can improve the models almost automatically on their own. The people using it define what good output looks like for their organization, and the platform turns that into model improvements. Co-founder Arjun Karanam said the company built an interface meant to make post-training approachable for nontechnical customers.
Who uses it
Harvey and Rogo already use the platform to post-train open-source models, for example on reviewing a particular type of legal contract or valuing a company with a specific method. Decagon, Mercor and Nvidia are also customers. The platform is now publicly available.
Competition and safety framing
Trajectory names Thinking Machines Lab, maker of the fine-tuning platform Tinker, as a direct competitor. It also highlights Applied Compute, which was assessed at $3.25 billion last month, and Prime Intellect, a firm that assists businesses in developing and training their own agents.
Co-founder Michael Elabd said the approach resembles recursive self-improvement but keeps humans in the training loop, which he argues makes it safer than autonomous self-improvement.
What the source does not say
The article gives no pricing, plan structure or technical requirements for the public platform, and it offers no performance data. The claims about faster and cheaper specialized AI come from the company and have not been independently verified in the source.