Thinking Machines Lab · posted · checked
Software Engineer, Sandboxing
LinuxPythonRust$300k – $400kOn-site · New York and San Francisco
Who can apply
On-site · New York and San Francisco
Location: New York / San Francisco
ABOUT THINKING MACHINES
The mission of Thinking Machines is to build AI that extends human will and judgment. We are training frontier models with Inkling, developing Tinker to let people make models their own, and crafting interfaces that broaden human-AI communication. We believe the future worth building is human, and we're hiring people who want to build it.
ABOUT THE ROLE
Our models and agents increasingly need to run code, use tools, and take actions in the world — safely, reliably, and at scale. The Core Services team builds the sandboxing infrastructure that makes this possible: the isolated execution environments where models write and run code, browse, and interact with tools, both for our researchers during training and for external users building on Tinker.
We're hiring a software engineer to help design, build, and operate this sandboxing platform. You'll work on the systems that isolate and constrain untrusted, model-generated code, and that scale to support thousands of concurrent executions across the company. This is foundational infrastructure: every research experiment and every product surface that lets a model take action depends on it being fast, secure, and dependable.
WHAT YOU'LL DO
Design, build, and operate sandboxed execution environments for running untrusted, model-generated code and tool calls at scale.
Improve isolation boundaries using technologies such as containers, microVMs, or gVisor-style kernels, balancing security against startup latency and throughput.
Build the scheduling, resource-management, and lifecycle systems that provision, reuse, and tear down sandboxes efficiently under heavy concurrent load.
Partner with researchers and Tinker's product team to expose sandboxing primitives that are simple to use and hard to misuse.
Instrument sandboxes for observability and abuse detection, and respond to novel escape or exploitation attempts as they're discovered.
Own reliability and performance of the sandboxing platform end-to-end, from API design down to the underlying virtualization layer.