xAI's mission is to create AI systems that can accurately understand the universe and aid humanity in its pursuit of knowledge. The RL infra team builds an internal RL framework to enable large scale reasoning training to build next state of art flagship models.
Requirements
- Python / Rust / C++ / CUDA
- Strong knowledge in quantization and numerics in LLM inference and training
- Experience in building testing / tools to analyze numerics
- Broad understanding of training/inference frameworks and reinforcement learning techniques
Responsibilities
- Develop low-precision solutions, from kernel to algorithm, to enhance efficiency in RL training and inference of flagship models.
- Analyze numerical properties of flagship models and their training dynamics, and develop tools to enhance and monitor numerical stability.
- Co-design models, systems, and recipes for numerically stable and optimal training and inference.
Other
- All engineers are expected to have strong communication skills.
- They should be able to concisely and accurately share knowledge with their teammates.
- Candidates are expected to be located near Palo Alto or open to relocation.
- Our team usually works from the office 5 days a week but allow work-from-home days when required.
