A one-year-old AI startup just tripled its ambitions with fresh cash
Trajectory, a startup founded about a year ago, raised $45 million in a Series A round at a $300 million valuation. Sequoia led the round, with Nvidia, Bessemer Venture Partners, and Bain Capital Ventures also participating.
The 18-person company was started by three Stanford classmates who previously worked at Google DeepMind and Apple. Its pitch is simple: let businesses retrain AI models for specialized work without needing an internal research team or expensive outside consultants, often called forward-deployed engineers.
Why AI models keep needing human babysitters
Current AI models, whether from OpenAI, Anthropic, or open-source projects, share a basic flaw. They don’t learn from their mistakes unless someone retrains them.
That’s a real problem for industries like law and finance, where firms want models that act like seasoned professionals rather than smart generalists. Trajectory co-founder Ronak Malde compared foundational models to a highly capable PhD student who’s always starting a new job. The economy, he told TechCrunch, values someone who’s done the job for 30 years, not someone brilliant on day one.
How the platform actually works
Trajectory’s approach is built around continual learning. Businesses start with an open-source AI model, connect it to Trajectory’s platform, and let it gather feedback on how the model performs in real use.
Trajectory then uses that feedback to retrain the model so it improves at specialized tasks over time. Malde says this lets lawyers, doctors, and other specialists refine models themselves, nearly automatically, without technical expertise. That could make deploying specialized AI both faster and cheaper for companies that can’t afford a dedicated engineering team.
Harvey and Rogo are already using the platform to post-train open-source models, helping them handle tasks like reviewing specific legal contracts or valuing companies using particular methodologies. Decagon, Mercor, and Nvidia are also customers. Trajectory is now opening its post-training platform to the public.
The safety question behind the pitch
Trajectory’s method shares some DNA with recursive self-improvement, the idea that AI systems can upgrade their own capabilities without human input. Co-founder Michael Elabd argues that’s exactly the risk Trajectory avoids by keeping humans in the loop.
Fully autonomous self-improvement, he says, could lead to unintended or dangerous outcomes. Users decide what counts as good performance for their organization, then feed that judgment into the model’s training.
The company isn’t alone in this space. It names Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, along with Applied Compute (valued at $3.25 billion last month) and Prime Intellect as direct competitors. Malde believes Trajectory’s edge is its team, which also includes recruits from OpenAI, Stripe, and Netflix, plus designers focused on making the interface usable for nontechnical customers.
Worth Watching
- Can Trajectory’s human-in-the-loop model stay competitive against rivals racing toward more autonomous training?
- Will the influx of nontechnical users actually produce reliable improvements, or introduce new quality risks?
- How will Trajectory differentiate itself from well-funded competitors like Thinking Machines Lab and Applied Compute?
Source: TechCrunch