Trajectory opens its AI training platform
Trajectory launched its post-training platform and announced a $45m Series A led by Sequoia at a $300m valuation.
A capable general model still needs to learn what good work looks like inside a particular organisation. Trajectory is building tools to make that adaptation easier.
What happened
Trajectory publicly launched its post-training platform and announced a $45m Series A led by Sequoia at a $300m valuation. Nvidia, Bessemer Venture Partners and Bain Capital Ventures participated.
The platform uses production data and human feedback to improve models for specialised tasks. It is designed to let people with domain knowledge help shape model behaviour without having to assemble the entire training process themselves.
Why it matters
A lawyer, analyst or other specialist may be well placed to judge an answer, even if they are not a machine-learning researcher. Turning that judgement into useful training feedback is a separate technical and product problem.
Tools that connect those steps could help companies adapt models more systematically. The important question is whether the resulting models improve on the tasks that matter to the business.
The bigger picture
Post-training is becoming part of the software infrastructure around AI applications. Companies need ways to collect feedback, assess improvements and decide when a model is ready to use.
Trajectory’s public launch makes that workflow available as a product. Its value will depend on measurable task performance and the cost of producing those improvements, with people retaining a clear role in defining the desired behaviour.