Agent orchestration
How agents plan, split work, coordinate, and ask for help.
At Uraion, research, engineering, and product work happen together. We build AI systems that can plan, remember, use tools, and check their work.
We do not have open roles right now. If your work is closely related to ours, you can still introduce yourself.
An idea matters when it works in a real system, under real constraints.
We define what we want to learn and how we will measure it.
We put the idea into working software and test it on real tasks.
Results and failures guide the next round of research.
These are research areas, not open roles.
How agents plan, split work, coordinate, and ask for help.
How systems store and reuse useful experience.
How smaller models can do focused work well.
How to run capable models locally, privately, and at lower cost.
How to test whether an agent's work is actually correct.
How to turn research into software people can rely on.
A paper, codebase, model, evaluation, or clear technical write-up tells us more than a polished résumé.
You know at least one technical area deeply.
You see the tradeoffs and failure modes around the work.
You test your claims and change your mind when the evidence changes.
You can turn an unclear goal into concrete work.
We do not keep a general hiring waitlist. If your work is a strong match, send a short note and one or two examples.