Test-time learning

Our system has already produced results beyond the reach of any human team.

It is one machine — post-training, agentic architectures, the data and evaluation layers beneath them — and we hire people who can build any part of it. The work is whatever the instrument needs next.

Research and engineering are one job here.

We do not separate research from engineering, and we do not carve the system into territories. You run whole loops — data, training, evaluation, analysis — and you ship.

Range, and proof.

• Strong machine-learning intuition and serious engineering fundamentals.
• Hands-on experience with LLM systems — post-training, agents, or the infrastructure beneath them.
• The ability to learn a new layer of the stack in weeks, not quarters.
• Code others can reproduce.

We weigh demonstrated ability above pedigree.

The terms are plain.

• A full-time, salaried appointment, hybrid in the San Francisco Bay Area.
• $320,000 base salary, plus meaningful early-stage equity.
• Compute to match the ambition.
• Relocation support and visa sponsorship where needed.
• Medical, dental, vision, and 401(k).