🔁Trajectory Raises $45M at $300M Valuation to Automate AI Fine-Tuning

Fine-tuning your model just got a lot less painful

Trajectory just raised $45 million at a $300 million valuation, and it wants to kill the fine-tuning grind. The round was led by Sequoia with Nvidia, Bessemer Venture Partners, and Bain Capital Ventures participating. The company is also making its post-training platform publicly available today. Here's the pitch. Most models don't learn from their mistakes without another training run, and that run usually needs an internal research team or an expensive consultant. Trajectory lets you start with an open source model, collect feedback on how it performs, and feed that back in for retraining. If you've been stitching this together with scripts and spreadsheets, this is the part that gets automated. The customer list is the real signal. Harvey and Rogo already use the platform to post-train open source models, and Decagon, Mercor, and Nvidia are customers too. The team pulled researchers and product designers from OpenAI, Stripe, and Netflix, and the nontechnical interface lets lawyers or doctors improve models without writing training code. Trajectory calls its approach continual learning and says it's similar to recursive self-improvement, but keeps humans in the loop, which is the part that makes it safer. The catch? Thinking Machines Lab, Applied Compute, and Prime Intellect all chase the same problem. Watch whether Trajectory's team advantage holds up once those competitors ship.