US — TechCrunch
Trajectory raises $45M to make AI model training accessible to nontechnical experts
AI1 min read
Trajectory is a 1-year-old startup that raised $45 million to help businesses teach AI models to get better at specialized jobs. Its tool lets people who are not engineers improve AI models using feedback.
What happened
Trajectory, a startup founded about a year ago by three Stanford classmates, announced a $45 million Series A funding round. The round was led by Sequoia and valued the company at $300 million. Its platform is now publicly available. It helps businesses use feedback to keep improving AI models that are built for special fields like law and finance.
Key numbers and facts
- $45 million raised in a Series A round
- $300 million valuation
- 18 employees
- Investors include Sequoia, Nvidia, Bessemer Venture Partners, and Bain Capital Ventures
- Customers include Harvey, Rogo, Decagon, Mercor, and Nvidia
Why AI models need extra training
Today's AI models, such as those from OpenAI and Anthropic, do not learn from their mistakes unless they are trained again. Companies that want AI for jobs like law or finance usually change a general model to fit their needs. This is called fine-tuning. Doing it often means hiring a research team or paying expensive consultants, known as forward-deployed engineers (FDEs).
Why it matters
Trajectory wants to make this work faster and cheaper. Its co-founder compares a standard AI model to a very smart student who is always starting a new job. Trajectory aims to let AI keep learning on the job, like a worker with 30 years of experience. If it works, more companies could build useful AI for special tasks without large expert teams. The company also says keeping humans involved makes its method safer than AI that improves itself without checks.
What to watch
- How many businesses start using the public platform
- Whether non-technical experts, such as lawyers and doctors, can improve models on their own
- How Trajectory competes with Thinking Machines Lab, Applied Compute, and Prime Intellect
- Whether customers like Harvey and Rogo see clear gains in their AI tools
Glossary
- Trajectory
- A startup that helps businesses keep improving AI models by using feedback to retrain them.
- fine-tuning
- Extra training that adjusts a general AI model so it works better on one special task.
- forward-deployed engineers (FDEs)
- Expert consultants who work inside a client's company to set up and adapt technology.
- Thinking Machines Lab
- A company founded by former OpenAI CTO Mira Murati that makes Tinker, a fine-tuning platform.