San Francisco-based AI startup Trajectory has raised $45 million in a Series A funding round led by Sequoia Capital, with participation from NVIDIA, Bessemer Venture Partners and Bain Capital Ventures. The round valued the company at $300 million and brought its total funding to $60 million.
Announced on November 2, 2026, the financing coincided with the introduction of Trajectory’s platform for post-training artificial intelligence models using real-world production data.
Trajectory was co-founded by Ronak Malde, Michael Elabd and Arjun Karanam, three Stanford alumni with previous experience at Google DeepMind and Apple. The company develops infrastructure that helps businesses customize AI models for specialized tasks and improve their performance using feedback collected from deployed applications.
Its platform enables organizations to connect AI models to their software, capture information such as user corrections, edits and task outcomes, and use those signals for further model training. The approach is designed to make model customization more accessible to businesses without requiring them to maintain dedicated AI research teams.
Trajectory is working with companies including Harvey and Rogo on post-training models for legal and financial applications. Other companies associated with its platform include Clay, Decagon and Mercor.
Alongside its platform announcement, Trajectory introduced two research previews focused on continual learning.
The first, Continual Grading, uses examples of expert evaluations to generate grading criteria for additional tasks. It is intended to reduce the amount of manual evaluation required when training models for specialized applications.
The second, GS1 (Group Size 1), is a reinforcement learning algorithm designed to train models using individual task executions rather than requiring multiple executions of identical tasks. Both technologies were released as research previews.
The Series A follows Trajectory’s previously announced $15 million financing backed by investors including Conviction and Bessemer Venture Partners. The latest round brings the company’s disclosed funding to $60 million as it develops its platform for continual AI model training.
