Trajectory Raises $45M to Automate AI Post-Training
Trajectory raised $45M Series A led by Sequoia to automate AI model post-training, enabling continuous improvement for specialized tasks.
The big picture: Trajectory, a startup founded by three Stanford classmates with experience at Google DeepMind and Apple, has raised $45 million in a Series A round to automate the post-training process for AI models, allowing them to continuously improve at specialized tasks.
Why it matters:
- Current AI limitations: Existing AI models, whether proprietary or open-source, do not learn from their mistakes without further training, requiring expensive fine-tuning by internal research teams or consultants.
- Cost and accessibility: Trajectory’s platform aims to make deploying specialized AI faster and cheaper by enabling non-technical specialists, such as lawyers and doctors, to improve models themselves.
- Continuous improvement: The company’s approach, known as continual learning, allows models to adapt and perform more like experienced professionals by learning from real-world feedback.
How it works:
- Feedback loop: Businesses connect Trajectory’s platform to their applications, allowing it to collect feedback on how open-source AI models perform in specialized contexts.
- Automated retraining: The platform uses this collected information to retrain and refine the models, enhancing their accuracy and effectiveness for specific tasks over time.
- Human-in-the-loop: Trajectory emphasizes human involvement in the training process, which co-founder Michael Elabd argues makes the approach safer compared to fully autonomous self-improvement.
The catch: Trajectory operates in a competitive market, facing rivals like Thinking Machines Lab, Applied Compute, and Prime Intellect, all working to automate AI post-training. The company believes its advantage lies in its experienced team, which includes researchers and product designers from Google DeepMind, Apple, OpenAI, Stripe, and Netflix, and its user-friendly interface designed for non-technical customers.
Key Facts
- Company: Trajectory
- Amount: $45M
- Round: Series A
- Investors: Sequoia (lead), Nvidia, Bessemer Venture Partners, Bain Capital Ventures
- Founders: Ronak Malde, Michael Elabd, Arjun Karanam
- Valuation: $300,000,000
- Announced: 2026-11-02
- Sector: AI Model Post-Training
AIPressRoom Weekly
Every round we reported this week, the people behind the companies, and the events coming up. One email, every Tuesday.