- Distributed systems, pipelines, perf for long-horizon agent runs
- High-bar-team pedigree, or early-at-a-now-grown-startup infra
The platform team builds the systems that make the whole motion possible: how agents run, how experts review their work, and how results feed future datasets and deployments. Ambitus lives or dies on how reusable its infrastructure is across contracts, so every contract should get easier, especially on the eval side. The engineer works on distributed systems, data pipelines, and performance optimization, plus the tools experts use to write and review tasks. Agent runs can take hours and depend on model APIs, remote environments, and background services, so the job is to make those workflows reliable, help the team understand failures, and keep performance up as usage grows (the JD frames the goal as supporting trillions of tokens of processing per week). Sandbox infra reliability is a live pain point right now.
Unlike the other roles, the platform hire is where Ambitus wants to be thoughtful rather than blitz. Leaning toward someone from a team with exceptional engineering standards, because startups often hand-wave good practices on the infra side and just ship, and platform is exactly where he wants high standards instead. A strong alternative is someone who was very early at a now-grown startup and had to stand up and scale infrastructure themselves.
What You’ll Own
- Long-horizon agent workflows (goal: trillions of tokens of processing per week)
- Services and data pipelines for task execution, evaluation results, and dataset delivery
- Investigating production issues; improving performance, resource usage, and recovery from failure
- Reliability of agent runs that span hours and depend on APIs, remote environments, and background services
- The internal tools experts use to write and review tasks, reducing repetitive manual work
Requirements
- Production software experience with a strong grounding in distributed systems, data pipelines, or performance optimization
- Sound engineering judgment; can investigate unfamiliar problems across a codebase
- Has taken a project from requirement through implementation and release
- Bonus: agent workflows, model inference, or evaluation systems
Signal
- High engineering standards from an exceptional team is the lean here (over scrappy zero-to-one), OR someone very early at a now-grown startup who stood up and scaled its infrastructure
- Brand-name / high-caliber-team signal matters
- Side projects and demonstrated agency
Who Will Thrive Here
- A distributed-systems or infra engineer from a high-bar team who wants to set real standards, not hand-wave them
- Someone who was early at a startup and felt the pain of scaling infra firsthand
- An engineer who makes long-running, failure-prone agent workflows reliable and fast
- High agency; thoughtful and standards-driven rather than purely blitz-scaling