Trajectory raises $45M to make AI model training accessible to nontechnical experts

Trajectory just bagged $45M at a $300M valuation to let non-engineers train AI models. They're betting that the real value in AI isn't just the base model, but the niche expertise you can bake into it without a PhD.

The era of massive, general-purpose LLMs is hitting a ceiling where specialized knowledge wins. If builders can fine-tune models using domain experts instead of expensive ML engineers, the barrier to building vertical AI drops overnight.

This accelerates the shift from horizontal AI giants to verticalized, niche-specific intelligence. It puts pressure on foundational model providers to simplify their own fine-tuning or watch specialized players eat their lunch.

• →Vertical SaaS founders can build expert-in-the-loop products faster, focusing on domain data rather than training infrastructure.
• →Consultancy models can emerge around data hygiene for AI, helping non-technical clients prep their proprietary knowledge for training.
• →Target the last mile of specialized industries like law, medicine, or construction where generic models fail the accuracy test.

• →The garbage in, garbage out trap: non-technical users might accidentally bake biases or messy data into their models, making them useless or dangerous.
• →Foundational providers like OpenAI or Anthropic could Sherlock this by making their own fine-tuning tools so intuitive that Trajectory becomes a feature, not a company.