Research, Tinker, RL Systems
Core
Building training systems for Tinker, including RL systems, numerics, and kernels to enable frontier model customization.
Role type
Senior IC machine-learning engineer (post-training/RL systems)
Builds
Post-training engine and RL training systems for Tinker
Domain
AI/ML infrastructure, large language model training
Deliverable
production ML models
Required skills
Python, deep learning frameworks (PyTorch, TensorFlow, JAX), distributed training debugging, RL algorithm design, system optimization
Preferred skills
Probability and statistics, RL training stability techniques, low-precision training and quantization, LLM serving stacks, scaling studies, open-source contributions
Technologies
PyTorch, TensorFlow, JAX, SGLang, vLLM
Responsibilities
Develop frontier customization techniques, co-design RL algorithms across the whole stack, debug RL runs, optimize post-training pipelines
Seniority
Senior, hands-on IC