Trajectory, a startup founded by former researchers from Google DeepMind, OpenAI and Apple, launched a self-service platform on November 2, 2026 that enables companies to fine-tune AI models using production data. Its software development kit collects edits and corrections made by users to AI-generated responses, then turns them into training material through weekly or more frequent post-training cycles. The platform includes continual grading, which converts human preferences into repeatable scores, and the GS1 algorithm, which is designed to learn from user actions in real time. Trajectory says it complies with SOC 2 standards and places customer data control at the core of its operations. The company, whose customers include Clay, Harvey, Decagon, Mercor and Rogo, raised $15 million in seed funding before attracting approximately $60 million to $65 million in total, including a Sequoia-led round that valued it at $300 million.
Source: Read the original article
