Research, Tinker, RL Systems

Thinking Machines Lab•San Francisco, CA

1d•$350,000 - $475,000•Onsite

About The Position

Thinking Machines is building AI that extends human will and judgment. They are training frontier models with Inkling, developing Tinker to let people make models their own, and crafting interfaces that broaden human-AI communication. The role focuses on building training systems for Tinker, including RL systems, numerics, kernels, and beyond. Tinker is a fine-tuning API that empowers researchers and developers to customize frontier AI to their needs, managing the infrastructure while allowing users flexibility in training models with their own data, algorithms, and for their own needs.

Requirements

Bachelor’s degree or equivalent experience in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline with strong theoretical and empirical grounding.
Proficiency in Python and familiarity with deep learning frameworks (e.g., PyTorch, TensorFlow, or JAX).
Comfort debugging distributed training and writing code that scales.
Clarity in communication, an ability to explain complex technical concepts in writing.
Strong interest in working on Tinker and increasing usefulness and adoption.

Nice To Haves

A strong grasp of probability, statistics, and ML fundamentals. Ability to distinguish between real effects, noise, and bugs from experimental data.
Experience with RL training stability techniques for large runs.
Familiarity with low-precision training and inference: numerics, quantization, and their implications for RL.
Hands-on work with LLM serving stacks (e.g., SGLang, vLLM, TokenSpeed, or custom engines).
Experience with scaling studies for large models.
Contributions to open-source training or inference frameworks.
PhD in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline with strong theoretical and empirical grounding; or, equivalent industry research experience.

Responsibilities

Develop frontier customization techniques and help build the best post-training engine in the industry.
Engage directly with researchers and companies pushing Tinker to its limits.
Work with internal research teams as well as contribute to open science and external partners.
Co-design RL algorithms and training systems across the whole stack, from RL science down to numerics and kernels, to enable anyone to post-train frontier models.
Debug RL runs in the wild, optimize post-training pipelines, and help users reach frontier-level results.