Trajectory, an AI infrastructure startup, has raised $40 million in new funding at a $300 million post-money valuation. Sequoia Capital led the round, with participation from Nvidia and Bessemer Venture Partners.

The raise comes just two months after Trajectory announced its $15 million seed round at a $115 million post-money valuation — a rapid jump in valuation for a company still in its first year.

Why Now

As closed-source AI models have grown more expensive, businesses have increasingly looked for ways to customize open-source models for specific use cases, while also improving the software "harnesses" that let models use tools and take actions. Trajectory sits at the intersection of both trends, working to make AI systems more useful, accurate, and cost-efficient for its customers.

What Trajectory Does

Trajectory is building a more automated, user-friendly way for businesses to customize open-source AI — without needing large teams of consultants or deep in-house technical expertise. Its techniques include self-distillation policy optimization, which uses feedback from real-world interactions to improve an AI agent's future decisions. For example, when a support agent gets stuck and a human resolves the case by consulting an additional database, that feedback loop helps the AI avoid the same mistake going forward.

The company also works with closed-source models, helping customers tune their AI harnesses to improve accuracy or lower operating costs. According to Trajectory's website, its customers include AI sales startup Clay, AI customer-support startup Decagon, and AI legal company Harvey.

The Team

Trajectory was founded in May 2026 by former Google DeepMind researchers Ronak Malde and Michael Elabd, along with former Apple researcher Arjun Karanam. The company currently has 12 employees. Its earlier seed round included backing from Conviction, Bessemer, and Radical Ventures.

Sources: Dealroom and The Information. Read the full reporting at the links above.