Trajectory's $45 million Series A prices its post-training platform at $300 million
Sequoia led a $45 million Series A for Trajectory, an 18-person startup that wants lawyers and finance desks to retrain their own models without hiring forward-deployed engineers.
| Trajectory | |
|---|---|
| Round | Series A |
| Amount | $45 million |
| Valuation | $300 million |
| Lead investor | Sequoia |
| Headcount | 18 |
The platform is now publicly available, which is the concrete change behind the round.
The round brought in Nvidia, Bessemer Venture Partners, and Bain Capital Ventures alongside Sequoia. The post-money valuation is $300 million.
Three Stanford classmates with stints at Google DeepMind and Apple founded the company about a year ago.
The buyer is a law firm or a finance desk that has been paying forward-deployed engineers to retrain models on its behalf. Trajectory's pitch is that the specialist replaces the FDE.
The document publishes no pricing, no latency figure, no model-size constraint, and no SLA. The architecture is described in one sentence: collect feedback, retrain.
Applied Compute closed at $3.25 billion last month, and Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, is also in this space. The $300 million valuation prices a market that a competitor already occupies at ten times higher.
Co-founder Michael Elabd argues that keeping humans in the training loop distinguishes Trajectory's approach from recursive self-improvement.
Harvey, Rogo, Decagon, Mercor, and Nvidia are named as current customers.
The round is real and the customers are named, but the architecture is one sentence long and the competitive gap is an order of magnitude. The public launch is the news; the funding is the confirmation.
After x.com. We did not report this. The pictures, if any, are theirs.
