Trajectory raises $45 million to let experts post-train AI

The round: Trajectory says it has raised a “$45M Series A led by Sequoia at a $300M valuation, with participation from NVIDIA, Bessemer, and Bain Capital Ventures.” The company announced the round on Nov. 2 in a post signed by “The Trajectory Team,” alongside the launch of what it calls “the Trajectory platform, the easiest and most intuitive platform for post-training on production data.”

The pitch: Trajectory argues that “a handful of research labs decided how AI behaves” and that everyone else “rented the result.” Its case is that people with subject-matter expertise, such as “the lawyer drafting a merger agreement, the banker building a model, the engineer shipping production code,” hold “the most valuable training signal there is.” The company also points to cost: “Every company building on AI is facing climbing token bills, and models still miss critical context,” it says, adding that post-training “often takes months of R&D and complex infrastructure.”

What it does: Post-training is the stage after a model’s initial training, when it is tuned for specific tasks. Trajectory says customers “plug in a set of tasks with a way to score them” and can then “train, evaluate and deploy” their own model on one platform. It describes the product as “the first self-serve product for post-training on production data” and says it replaces “four products (sandboxes, evals, training and deployment)” with “one product and about five lines of code.” The company promises “one click to train, one click to deploy,” without customers managing sandboxes, orchestration or infrastructure.

Two previews: Trajectory lists two features as research previews. Continual Grading, it says, “learns from a few examples of your judgment and applies those standards across your production data, reducing the need for manual labeling.” Group Size 1 “lets you train on real production interactions, so your model can keep improving after it ships.” The company calls them “the first of many releases towards continual learning: AI that improves from the work it does, the way people do.”

Customers and results: Trajectory says “builders at Harvey, Rogo, Mercor and NVIDIA are already training on Trajectory,” and that one unnamed customer “trained more than eight models for complex, long-horizon agentic tasks in just two weeks.” Working with Nvidia and Harvey, the company says it post-trained Nemotron 3 Ultra on frontier legal tasks “within 24 hours of its release.” It also says it took a Nemotron Lightning model “from 0% to an 8.3% all-pass rate on Harvey’s LAB benchmark, matching Opus 5.5’s score at 100x lower cost and 3.8x faster speed.” Opus 5.5 is one of Anthropic’s Claude models. Those figures are Trajectory’s own.

The team: Trajectory says its research team “contributed to Gemini 3 at DeepMind, GPT-4o at OpenAI, and Muse infrastructure at Meta Superintelligence,” and that its product team “built and launched unique interfaces at Stripe, Netflix, and Apple.”

Our read: Trajectory is betting that the next gains in useful AI come less from bigger base models and more from tuning existing ones on real work, graded by the people who do that work. Its pitch is aimed squarely at companies that already run AI in production and pay for it by the token: if a cheaper tuned model can match a frontier model on a narrow task, the savings add up. The Harvey benchmark result is the company’s strongest evidence, but it comes from a single benchmark, with the results reported by Trajectory itself. The research-preview labels on Continual Grading and Group Size 1 matter too: the continual-learning features at the center of the pitch are early. With Sequoia leading and Nvidia both investing and listed as a customer, the round gives the company room to prove those claims outside its own posts.

What’s next: Trajectory says the platform is self-serve and that users can “start training your own model today.” The announcement does not list pricing. Watch for customer-published results beyond Trajectory’s own figures and for the research previews to move to general release.

Sources: Trajectory: Own Your Value Function: Introducing the Trajectory Platform (Nov. 2, 2026)