Trajectory Secures $45M Series A Led by Sequoia to Help AI Models Learn From Feedback
By James Dargan · · 1 min read
Trajectory, a startup automating the post-training of AI models for specialized work, has secured $45 million in a Series A round at a $300 million valuation. Sequoia led the round, with participation from Nvidia, Bessemer Venture Partners and Bain Capital Ventures.
Founded about a year ago by Stanford classmates Ronak Malde, Michael Elabd and Arjun Karanam, the 18-person company targets a key limitation of today’s AI models: they cannot learn from mistakes without further training. Businesses typically address this through costly in-house research teams or forward-deployed engineers.
Trajectory uses continual learning, connecting to a company’s applications to collect feedback on an open source model’s performance and retrain it over time. Malde told TechCrunch that foundation models resemble a brilliant PhD student starting a new job, while the economy values decades of on-the-job experience, and said specialists can now improve models almost automatically.
Harvey and Rogo already use the platform to help models perform like experienced lawyers and financial analysts, alongside customers including Decagon, Mercor and Nvidia. The platform is now publicly available.
Elabd argued that keeping humans in the loop makes the approach safer than autonomous self-improvement. Trajectory competes with Thinking Machines Lab, Applied Compute and Prime Intellect.
Featured image: Credit: Trajectory
