US startup Trajectory has raised $45 million in a Series A round to expand an AI model training platform built for nontechnical experts, and is now valued at $300 million. Sequoia led the round, with participation from Nvidia, Bessemer Venture Partners and Bain Capital Ventures.
The roughly one-year-old, 18-person company was founded by three Stanford classmates—Ronak Malde, Michael Elabd and Arjun Karanam—who previously worked at Google DeepMind and Apple. Their pitch sits at a practical problem: large general-purpose models do not keep learning from mistakes unless they are trained further, and many firms still hire costly research teams or consultants known as forward-deployed engineers to fine-tune systems for law, finance and similar fields.
How the AI model training platform works
Trajectory focuses on continual learning, a method in which a model keeps improving from new feedback rather than remaining fixed after one training pass. Businesses can start with an open-source model, connect the platform to their applications, and let it gather signals on how the model performs. Trajectory then retrains the system so it gets better at specialised tasks over time.
Malde compared foundational models to a very smart PhD graduate who is always starting a new job, adding that “what the economy values is someone who worked on the job for 30 years.” With the platform, he said, lawyers, doctors and other specialists can improve models “near automatically” themselves, which could make specialised AI faster and cheaper to deploy. Users decide “what good looks like for their organization and can turn that into improvements for their model,” Malde said.
Who is already using it
Companies such as Harvey and Rogo are already using Trajectory to post-train open-source models so they behave more like experienced lawyers or financial analysts—for example by learning from past contract reviews or company valuations that follow a set method. Other customers named by the company include Decagon, Mercor and Nvidia. Trajectory is now making this post-training product publicly available.
Elabd has argued that keeping humans in the training loop is safer than fully autonomous recursive self-improvement, in which AI systems try to upgrade themselves without oversight. Karanam has stressed product design: the firm has built an interface meant to make post-training usable for non-technical customers, not only for machine-learning staff.
Funding, team and competition
Beyond the $45 million raise and $300 million valuation, Trajectory says its edge is the team. Malde was among researchers DeepMind hired from Windsurf; Elabd also worked at DeepMind; Karanam worked on generative foundation models at Apple. The startup has also hired researchers and product designers from OpenAI, Stripe and Netflix.
It faces peers including Thinking Machines Lab—linked to former OpenAI CTO Mira Murati and the fine-tuning product Tinker—plus Applied Compute, which was valued at $3.25 billion last month, and Prime Intellect, which helps firms build and train their own AI agents. Malde still frames Trajectory’s main advantage as people and accessibility rather than a closed technical monopoly.
Aasan alfaaz mein
Trajectory naam ki company ne AI seekhne wala platform banane ke liye 45 million dollar uthaye hain. Ab iski qeemat lagbhag 300 million dollar batai ja rahi hai. Lawyer, doctor aur doosre experts is se models ko feedback de kar behtar bana sakte hain, bina bari research team ke. Jab yeh tools aam logon tak pahunchain, specialised AI sasta aur tez ho sakta hai.
