Trajectory has raised $45 million in a Series A round at a $300 million valuation, as it works to make specialized AI model training easier for people without technical expertise. Sequoia led the round, with participation from Nvidia, Bessemer Venture Partners and Bain Capital Ventures.
Founded about a year ago by three Stanford classmates, Trajectory aims to help AI models improve at specialized tasks by learning from feedback. The company has 18 employees, and its founders have backgrounds at Google DeepMind and Apple.
Businesses often adapt general-purpose AI models for fields such as law and finance through fine-tuning. That work can require in-house research teams or expensive consultants. Trajectory’s platform instead connects an open-source model to a company’s applications, collects feedback on its performance and uses that information to retrain the model over time.
Co-founder Ronak Malde said the platform is intended to let specialists, including lawyers and doctors, shape how a model performs without relying on a technical team to manage the process. Companies can define what successful work looks like for their organization and use that standard to guide improvements.
Harvey and Rogo use Trajectory to post-train open-source models for tasks such as reviewing specific types of legal contracts or valuing companies according to particular methods. Other customers include Decagon, Mercor and Nvidia. The startup is now making its post-training platform publicly available.
Trajectory’s approach involves human feedback, which co-founder Michael Elabd said distinguishes it from autonomous self-improvement systems that could produce unintended outcomes. The startup faces competition from Thinking Machines Lab, Applied Compute and Prime Intellect, which also work on AI training or post-training tools.
The founders say the product’s interface is central to making post-training accessible to nontechnical users. The team also includes researchers and product designers previously employed by OpenAI, Stripe and Netflix.
Sources: TechCrunch Startups.