The most expensive sentence in enterprise AI right now is “we need engineers for that.” Trajectory, a startup founded about a year ago by three Stanford classmates, has raised a $45 million Series A led by Sequoia to automate the post-training of AI models, so lawyers, doctors, and other specialists can improve the models they use without hiring research teams.
Key takeaways
- Trajectory raised $45 million in a Series A at a $300 million valuation, led by Sequoia with participation from Nvidia, Bessemer Venture Partners, and Bain Capital Ventures, per TechCrunch.
- The platform automates post-training: businesses start from an open-source model, connect it to their applications, collect feedback on real use, and retrain the model automatically.
- Customers already include Harvey and Rogo, which use it to post-train open-source models for legal and financial work, alongside Decagon, Mercor, and Nvidia. The platform is now opening to the public.
- The 18-person company was founded by CEO Ronak Malde, Michael Elabd, and Arjun Karanam, three Stanford classmates with experience at Google DeepMind and Apple. It raised a $15 million seed in May 2026.
The problem: fine-tuning is gated on scarce engineers
Current AI models share a structural limitation: they do not learn from their mistakes without further training. Businesses adapting general models for specialized fields like law and finance typically get around this by fine-tuning, which usually means hiring internal research teams or paying for forward-deployed engineers, the expensive consultants who embed with customers to make models actually work.
Trajectory’s bet is that this gate can be automated away. Its platform connects an open-source model to a company’s applications, collects feedback on how the model performs in real use, and uses that feedback to retrain the model, helping it improve at specialized tasks over time. “With Trajectory, lawyers, doctors, and other specialists can improve the models near automatically themselves,” Malde told TechCrunch. The people using the platform decide “what good looks like for their organization and can turn that into improvements for their model.”
Traction and the team behind it
The customer list is unusually concrete for an 18-person company. Harvey and Rogo already use Trajectory to post-train open-source models so they perform more like experienced lawyers or financial analysts, learning from previous tasks such as reviewing a particular type of legal contract or valuing a company using a specific methodology. Decagon, Mercor, and Nvidia are also customers, and the public launch coincides with the funding announcement.
The founding team’s credentials matter in this market. Malde was among the researchers Google DeepMind acqui-hired from Windsurf; Elabd also worked at DeepMind; Karanam previously worked on generative foundation models at Apple. The company has recruited researchers and product designers from OpenAI, Stripe, and Netflix, and Karanam says the interface itself is part of the product: post-training has to be accessible to nontechnical customers, not just possible for them.
Notably, co-founder Michael Elabd draws a line between Trajectory’s approach and recursive self-improvement, the idea of AI systems improving themselves autonomously. Keeping humans in the training loop, he argues, is what makes continual learning safe; autonomous self-improvement could lead to unintended or even dangerous outcomes.
The market forming around post-training
Trajectory is not alone. Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, built the fine-tuning platform Tinker; Applied Compute was valued at $3.25 billion last month; and Prime Intellect helps companies build and train their own AI agents. The pattern is the familiar one: as foundation models commoditize, the value moves to the layer that adapts them to specific work, and investors are pricing that layer aggressively. A $300 million valuation for a one-year-old, 18-person company is this year’s market speaking again: compressed timelines, proven teams, real contracts.
FAQ
What is Trajectory?
Trajectory is a startup that automates post-training of AI models: its platform connects an open-source model to a company’s applications, collects feedback on real use, and retrains the model so it improves at specialized tasks. It was founded about a year ago by three Stanford classmates and has 18 employees.
How much did Trajectory raise?
$45 million in a Series A at a $300 million valuation, led by Sequoia with Nvidia, Bessemer Venture Partners, and Bain Capital Ventures participating, per TechCrunch. The company previously raised a $15 million seed announced in May 2026.
What is AI post-training?
Post-training is everything that happens to a model after its initial training: fine-tuning it on specialized data, aligning it with feedback, and adapting it to particular tasks. It is how a general model becomes good at reviewing a specific kind of contract or valuing companies with a specific methodology.
Who founded Trajectory?
CEO Ronak Malde, Michael Elabd, and Arjun Karanam, three Stanford classmates. Malde and Elabd worked at Google DeepMind (Malde via the Windsurf acqui-hire), and Karanam worked on generative foundation models at Apple.
More from Grey Journal: Sebastian Thrun Raises $130 Million for Dulo, His Bet on AI That Designs Hardware.
