Trajectory raises $45M to automate AI model customization
Trajectory, a startup founded by Stanford alumni, secured $45 million in Series A funding at a $300 million valuation.
Trajectory has emerged from stealth with a $45 million Series A funding round at a $300 million valuation. The round was led by Sequoia Capital, with participation from Nvidia, Bessemer Venture Partners, and Bain Capital Ventures.
The startup, founded about a year ago by three Stanford classmates, aims to automate post-training AI model customization. Its platform allows businesses to connect an open-source model to their applications, collect feedback, and retrain the model without deep technical expertise.
Platform and capabilities
Trajectory offers a no-code interface for post-training through continual learning. Users can upload datasets, define training objectives, and deploy custom models via a visual workflow builder. The system is designed for domain experts in regulated fields like healthcare, legal, and finance, who have sensitive proprietary data that cannot be sent to third-party APIs due to compliance requirements like HIPAA.
Co-founder Arjun Karanam said the company developed an interface to make post-training more accessible to nontechnical customers. Engineers manage the retraining steps, enabling lawyers, doctors, and financial analysts to refine models within their own workflows. Companies including Harvey, Rogo, Decagon, Mercor, and Nvidia are using the platform to post-train open-source models, seeing performance that matches seasoned professionals.
Safety and human oversight
The platform emphasizes keeping humans in the loop during model improvement. Co-founder Michael Elabd argues that keeping people involved in training makes the approach safer than autonomous self-improvement. While Trajectory's approach shares similarities with recursive self-improvement-the idea that AI systems can improve their own capabilities-it maintains human oversight as a core safety differentiator.
Market and competition
Trajectory operates in a growing market for AI customization tools, facing competition from several other startups and established players. The competitive landscape includes companies offering programmatic access to model customization.
Foundation models often cannot adjust after deployment without additional training, forcing businesses to either fine-tune models internally or pay costly consultants. Trajectory positions itself as making this customization accessible to non-engineers.
Team and vision
The founding team combines experience from leading AI labs and tech companies. Founders include former DeepMind researcher Ronak Malde, and Michael Elabd and Arjun Karanam, who have experience from Apple and Google. The 18-person team has also recruited researchers and product designers from OpenAI, Stripe, and Netflix. Funding will be used for product development and expansion.
Co-founder Ronak Malde stated that lawyers, doctors and other specialists can use the platform to improve models almost automatically. He added, "But what the economy values is someone who worked on the job for 30 years." The vision is to lower barriers to specialized AI adoption and accelerate industry transformation through user-driven customization. Trajectory is now making its post-training platform publicly available to businesses.
