Building the simulation layer that teaches AI to reason before it acts.
Member of Technical Staff, Software Engineer
We think agents fail silently in production because they never trained in environments that actually resemble the real world. This team is constructing those environments—high-fidelity, composable sandboxes where frontier models can practice tool use, multi-step planning, and failure recovery long before they touch a customer. It is a genuinely unsolved systems problem with implications for every enterprise shipping agentic software. You will own full-stack features end-to-end: designing the interactive surfaces where engineers configure simulations, wiring up the backend orchestration that spins up isolated sandboxes at scale, and shaping the APIs that let training pipelines treat our platform as programmable infrastructure. The work sits at the intersection of developer tooling, distributed systems, and the operational guts of modern AI.
What they're looking for
- Fluency across the stack with TypeScript, React, Python, and PostgreSQL—you can debug a sluggish query and refine a component's interaction state in the same afternoon.
- Comfort designing and consuming REST APIs with an instinct for clean boundaries between services.
- 0–3 years of professional software experience, including internships or substantial open-source work.
- Clear, unhurried communication: you write tight design docs and leave code reviews better than you found them.
- Energized by working five days a week in San Francisco alongside a small, focused team that moves with urgency but without panic.