Software Engineer, Applied AI - Environments Platform

Frontier labs train and evaluate their models inside environments Mercor builds: sandboxed codebases, enterprise applications, multi-tool agent worlds, and the verifiers that score what an agent did. The Environments Infrastructure team owns the platform that builds those environments, runs them at scale, and gets them to the customer, however a given customer wants to consume them, from a service they call directly to a dataset they run in their own stack.

That makes this team the layer between everything Mercor authors and everyone who runs it. We build the core environment runtime, the structure of the environments themselves, the services that other teams at Mercor build on, the integrations our lab customers build against, and the pipelines that carry an environment from authoring to a training run. Every environment Mercor ships goes through this stack.

The team is small, the surface area is large, and most of the big design questions are still open. You will own real pieces of this platform end to end and see what you build in use at the biggest AI labs within weeks of shipping.