Trajectory, a San Francisco AI infrastructure startup founded by former Google DeepMind and Apple researchers, launched a self-service platform on November 2, 2026 that enables companies to update their models using user corrections. The company uses a technique called self-distillation policy optimization to continuously refine models from real-world edits, potentially as often as weekly. Trajectory emerged from stealth in May 2026 with a $15 million seed round at a $115 million post-money valuation, then raised a $40 million Series A at a $300 million valuation in August. Its disclosed funding now totals approximately $55 million to $65 million, with Sequoia Capital, Nvidia and Bessemer Venture Partners among its backers. Early users include Clay, Decagon and Harvey, while the automated approach could reduce reliance on costly fine-tuning consultants but raises concerns about the quality of user feedback and control over regulated applications.
Source: Read the original article
