Research, Tinker, RL Systems

Thinking Machines Lab•San Francisco, CA

5d•Onsite

About The Position

Thinking Machines is building AI intended to extend human will and judgment. Its work includes training frontier models with Inkling, developing Tinker to help people customize models, and creating interfaces for broader human-AI communication. Tinker is a fine-tuning API that lets researchers and developers customize frontier AI using their own data, algorithms, and needs, while Thinking Machines manages the infrastructure. This role focuses on building Tinker's training systems, including reinforcement learning systems, numerics, and kernels. The role involves developing frontier customization techniques and helping build a leading post-training engine. It calls for a whole-stack understanding of recipes, data pipelines, and training systems, and includes working directly with researchers and companies using Tinker. The position contributes to internal research, open science, and external partnerships.

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 such as PyTorch, TensorFlow, or JAX.
Comfort debugging distributed training and writing code that scales.
Clear communication skills, including the ability to explain complex technical concepts in writing.
Strong interest in working on Tinker and increasing its usefulness and adoption.

Nice To Haves

Strong grasp of probability, statistics, and machine learning fundamentals.
Ability to distinguish real effects, noise, and bugs in experimental data.
Experience with reinforcement learning training stability techniques for large runs.
Familiarity with low-precision training and inference, including numerics, quantization, and their implications for reinforcement learning.
Hands-on experience with LLM serving stacks such as 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 post-training engine for Tinker.
Co-design reinforcement learning algorithms and training systems across the stack, from RL science to numerics and kernels.
Debug reinforcement learning runs and optimize post-training pipelines.
Help Tinker users achieve frontier-level results.
Work with internal research teams and contribute to open science and external partnerships.