Trajectory Raises $45 Million in Series A Funding for AI Training
Artificial intelligence startup Trajectory announced that it secured $45 million in a Series A funding round at a $300 million valuation. The investment was led by Sequoia, with participation from Nvidia, Bessemer Venture Partners, and Bain Capital Ventures. Founded about a year ago by three Stanford classmates with background experience at Google DeepMind and Apple, the 18-person company aims to automate the process of fine-tuning foundational models for specialized industries.
Current artificial intelligence models developed by companies like OpenAI and Anthropic share a critical limitation because they do not learn from their mistakes without further training. Businesses adapting these general models for specialized fields such as law and finance typically get around this by hiring internal research teams or relying on expensive consultants known as forward-deployed engineers. Trajectory wants to eliminate this hurdle by allowing models to learn from feedback and continuously improve at specialized tasks.
How Trajectory Approaches Continual Learning and Post-Training
The startup uses an approach known as continual learning to help organizations improve foundational technology. Businesses using Trajectory start with an open-source model and connect the platform to their applications. This connection allows the system to collect feedback on how the model performs on specific tasks. Trajectory then uses that information to retrain the model, helping it improve over time without requiring heavy technical intervention.
Trajectory co-founder Ronak Malde compared foundational models to a very smart PhD student who is always starting a new job. Malde told reporters that what the economy values is someone who worked on the job for 30 years. With Trajectory, lawyers, doctors, and other specialists can improve the models near automatically themselves, making deploying specialized artificial intelligence faster and cheaper.
Real-World Applications in Law, Finance, and Customer Support
Several prominent companies are already using Trajectory's platform to post-train open-source models. Firms like Harvey and Rogo use the software to help models perform more like experienced lawyers or financial analysts by learning from previous tasks. These tasks include reviewing a particular type of legal contract or valuing a company using a specific methodology. Trajectory counts Decagon, Mercor, and Nvidia among its other customers, and the startup is now making its post-training platform publicly available.
The people using Trajectory decide what good looks like for their organization and can turn that into improvements for their model, according to Malde. This setup allows non-technical experts to define quality metrics and feed human guidance back into the system. Arjun Karanam, one of the co-founders, noted that the startup has developed an interface that makes post-training more accessible to these non-technical customers by prioritizing design.
Safety, Human Oversight, and Recursive Self-Improvement
Trajectory's approach to continual learning shares some similarities with recursive self-improvement, the idea that artificial intelligence systems can improve their own capabilities. However, co-founder Michael Elabd argues that keeping humans involved in the training process makes Trajectory's approach safer. Autonomous self-improvement could lead to unintended or even dangerous outcomes, making human guidance a necessary component of ongoing model refinement.
The founding team brings substantial technical pedigree to the venture. Malde was among the handful of researchers Google DeepMind acqui-hired from Windsurf, and Elabd also worked at DeepMind. Karanam previously worked on generative foundation models at Apple. Furthermore, the startup has recruited researchers and product designers from OpenAI, Stripe, and Netflix to build its team.
The Competitive Landscape in Automated AI Training
Trajectory is not the only startup trying to automate post-training for enterprise clients. The company considers Thinking Machines Lab a direct competitor. Thinking Machines Lab was founded by former OpenAI CTO Mira Murati and is the company behind the fine-tuning platform Tinker. Other competitors include Applied Compute, which was valued at $3.25 billion, and Prime Intellect, which helps companies build and train their own artificial intelligence agents.
Despite this competition, Malde believes the startup's biggest advantage is its team and product focus. By lowering the barrier to entry for fine-tuning open-source models, Trajectory hopes to capture a large share of the growing market for customized enterprise solutions. The newly acquired capital will support the public rollout and expansion of the platform.
