Thinking Machines Lab · posted · checked
Software Engineer, Frameworks & Libraries
Python$300k – $400kOn-site · New York and San FranciscoMid-level
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
We're hiring a Software Engineer, Frameworks & Libraries to turn the functionality we build internally into the open-source core libraries that the community builds on. You will package the code that powers our model launches and the Tinker Cookbook into well-designed, well-documented libraries, so that when we release a new model, researchers and developers can sample from it, post-train it, and evaluate it from day one.
You will own the path from internal prototype to public library across the full stack of model-interaction primitives. Each layer builds on the one before it, so the interfaces you design need to compose cleanly, stay stable as models and training methods change, and hold up under real use by Tinker customers.
This is a technical-first role at the boundary between research, infrastructure, and our open-source community. It's ideal for someone who cares about API design, backwards compatibility, and developer ergonomics, and who is just as comfortable reading a tokenizer's chat template as they are shipping a release.
WHAT YOU'LL DO
Package internal functionality into external, open-source core libraries (e.g., tml-renderers) and maintain them with clear versioning, changelogs, deprecation policies, and release automation
Support model launches by ensuring renderers, tokenizers, parsers, and tool-call handling for new models ship correctly and are ready for sampling and post-training on day one
Design and build the layers of the stack that follow the renderer: context compaction, session management, the rollout loop and runtime, a rollout viewer, and graders
Contribute to the Tinker Cookbook by turning these libraries into reference recipes and examples that demonstrate best practices for supervised and reinforcement learning
Define clean, composable, typed APIs between libraries, and keep the Cookbook, the Tinker SDK, and our internal training and inference systems compatible with each other