Trajectory raises $45M to let legal and finance pros retrain AI without engineers
The startup aims to let legal and finance experts improve AI models through automated feedback loops without hiring specialized engineering teams.
Key points
- Trajectory secured $45 million in Series A funding at a $300 million valuation.
- The platform enables non-technical specialists to retrain models using application feedback.
- Early customers include Harvey, Rogo, Decagon, Mercor and Nvidia.
Investors also include Nvidia, Bessemer Venture Partners and Bain Capital Ventures.
What changed
The 18-person company is making its post-training platform publicly available. Previously, businesses adapting general AI models for specialized fields like law or finance relied on internal research teams or expensive forward-deployed engineers. Trajectory aims to remove that barrier by automating the fine-tuning process.
How it works
The platform uses an approach called continual learning. Users connect the system to their applications, allowing it to collect performance feedback. Trajectory then uses that data to retrain open-source models. This allows specialists to define what good performance looks like and turn those standards into model improvements over time.
What the evidence means
Current AI models do not learn from mistakes without further training. Trajectory’s method shares similarities with recursive self-improvement, where systems enhance their own capabilities. However, the startup explicitly states that specialists can improve models near automatically, suggesting a shift in who controls model refinement.
Cost and market position
Trajectory claims its capability could make deploying specialized AI faster and cheaper. Companies like Harvey and Rogo already use the platform to post-train models for legal and financial analysis tasks.
Why it matters
- Product teams can reduce dependency on external forward-deployed engineers for model customization.
- The $300 million valuation signals strong investor confidence in automated fine-ting tools.
- Public availability of the platform lowers the barrier for smaller firms to adopt specialized AI.
- Integration with existing applications allows for continuous, feedback-driven model improvement.
Topics: AI training, fine-tuning, startup funding, open-source models, continual learning
Updated
The strongest doubt
Counterpoint from the sources
The source notes that current AI models share a critical limitation: they do not learn from mistakes without further training.
For readers
Business
- Trajectory claims its tool makes deploying specialized AI faster and cheaper.
- Early customers include Harvey, Rogo, Decagon, Mercor and Nvidia.
Developers
- The platform allows connection to applications to collect performance feedback.
- Users can start with open-source models and retrain them automatically.
Investors
- Trajectory raised $45 million at a $300 million valuation.
- Sequoia led the round with participation from Nvidia and others.
Numbers in context
Circle area is proportional to the magnitude; labels retain signs. Figures may describe different measures. Read their contexts above.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
Continual learning
An approach where AI models improve over time using new feedback data.
Forward-deployed engineers
Expensive consultants hired to fine-tune general AI models for specific business needs.
Post-training
The process of refining an AI model after its initial development phase.
Questions readers ask
Who invested in Trajectory's latest funding round?
Sequoia led the round, with participation from Nvidia, Bessemer Venture Partners and Bain Capital Ventures.
Does Trajectory work with proprietary models?
According to the report, businesses can start with open-source AI models connected to their applications.
Connected stories
Reading receipt
- Trajectory secured $45 million in Series A funding at a $300 million valuation.
- The platform enables non-technical specialists to retrain models using application feedback.
- Early customers include Harvey, Rogo, Decagon, Mercor and Nvidia.
