Trajectory Secures Millions to Let Experts Train AI Without Engineers

Foundational artificial intelligence models act like brilliant PhD graduates starting a new job, possessing vast general knowledge but lacking the deep, practical experience the economy actually values. Businesses adapting these systems for specialized fields like law and finance usually rely on internal research teams or expensive forward-deployed engineers to constantly fine-tune the software.

Without that manual intervention, current models from major developers fail to learn from their mistakes or improve over time from daily feedback. Trajectory, an eighteen-person startup founded by former Google DeepMind and Apple researchers, wants to eliminate that technical barrier by automating the post-training process. The company recently raised forty-five million dollars in a Series A funding round at a three hundred million dollar valuation, led by Sequoia with participation from Nvidia, Bessemer Venture Partners, and Bain Capital Ventures.

By connecting the Trajectory platform to existing applications, businesses can automatically collect performance feedback and retrain open-source models without hiring dedicated engineering consultants. Our read is that the race to commercialize artificial intelligence is shifting away from building bigger foundational brains and toward building better, frictionless feedback loops for everyday professionals.

Co-founder Ronak Malde argues that specialists like lawyers and doctors should be able to define what good work looks like for their own organizations and automatically fold those lessons back into the software.

While fully autonomous self-improvement remains a risky proposition, Trajectory deliberately keeps humans inside the training loop to maintain safety while competing against heavily funded rivals in the automated post-training space.