The platform tackles a known limitation of AI models: they do not learn from their mistakes without further training. Businesses adapting general models for specialised fields such as law or finance usually fine-tune them, which often means hiring internal research teams or expensive consultants known as forward-deployed engineers. Trajectory automates that process. Customers start with an open source model, connect the platform to their applications, and it collects feedback on performance and retrains the model so it improves at specialised tasks over time. Co-founder Ronak Malde says lawyers, doctors and other specialists can improve the models near automatically themselves.

Customers including Harvey, Rogo, Decagon, Mercor and Nvidia already use the platform to post-train open source models, for example to review particular types of legal contract or value companies using a specific methodology. Trajectory is not alone in this space: it names Thinking Machines Lab, the company behind the Tinker fine-tuning platform, as a direct competitor, along with Applied Compute and Prime Intellect.

Co-founder Michael Elabd argues that keeping humans involved in the training makes Trajectory's approach to continual learning safer than fully autonomous self-improvement, which he suggests could lead to unintended outcomes. The company also credits its interface, designed to make post-training accessible to nontechnical customers, as a differentiator.