Tasks:
- * Build training benchmarks and CI regression systems
- * Build training, inference, trajectory management, and reward modules
- * Build troubleshooting and model delivery tools
- * Design and validate engineering solutions for training requirements
- * Develop and optimize reinforcement learning training frameworks
- * Develop experiment configuration and tracking tools
- * Ensure training numerical correctness and stability
- * Implement GRPO, PPO, and on-policy distillation workflows
- * Improve framework modularity and serviceability
- * Integrate and productionize new training methods
- * Integrate distributed training and inference backends
- * Profile and optimize end-to-end performance and resource utilization
- * Support embodied model reinforcement learning training workflows
Perks/Benefits:
Skills/Tech stack required:
[AReaL] [Asynchronous reinforcement learning] [Benchmarking] [CI/CD] [CUDA] [Distributed communication] [Distributed inference] [Distributed Training] [FSDP] [GPU Optimization] [GRPO] [Linux] [Machine Learning] [Megatron] [Memory Optimization] [Model Inference] [Model Training] [Multimodal Machine Learning] [Nsight] [On Policy] [On policy Distillation] [Performance Profiling] [Policy Distillation] [PPO] [Python] [PyTorch] [PyTorch Profiler] [Reinforcement Learning] [RLinf] [SGLang] [Triton] [VeRL] [VLLM]
Educational requirements:
N/A
Role(s):
[AI Infrastructure Engineer] [Engineer] [Infrastructure Engineer] [Reinforcement Learning Infrastructure Engineer]