Sequoia-backed Trajectory opens a self-serve platform for post-training AI models on real user corrections
An SDK captures edits and corrections in a customer's product and feeds weekly or faster post-training cycles; Crypto Briefing reports a Sequoia-led round at a $300 million valuation.
Why it matters
Companies building on AI models collect a constant stream of user fixes, but most of it is thrown away. Trajectory, founded by former Google DeepMind, OpenAI and Apple researchers, sells tooling to turn those corrections into regular model updates, and it has moved from working with a few partners to letting any qualifying business sign up. Its customers include Clay, Harvey, Decagon, Mercor and Rogo. Watch for independent results from its Continual Grading and GS1 research previews.
Key facts
- Platform went live on November 2, 2026
- Research previews: Continual Grading and GS1
- Seed round of $15 million at a $115 million valuation
- Sequoia-led round valued the company at $300 million
- Total investment approximately $60-65 million
- Customers include Clay, Harvey, Decagon, Mercor and Rogo