Member of Technical Staff (Research Engineer - LLM Systems & Performance)

Lead and optimize end-to-end LLM systems, from supervised fine-tuning and reinforcement learning pipelines to high-throughput production inference. Develop and enhance SFT/RL components (e.g., Verl, SkyRL), covering data loading, training loops, logging, and evaluation. Contribute to LLM inference infrastructure (e.g., vLLM, SGLang) with batching, KV-cache management, scheduling, and serving optimizations. Profile and optimize end-to-end performance (throughput, latency, memory bandwidth) with tools like Nsight and profilers to identify bottlenecks. Work across multi-GPU clusters using NCCL, NVLink and various parallelism strategies (data/tensor/pipeline/expert/context) and explore quantization (INT8/FP8/FP4, mixed precision). Collaborate with researchers to move ideas from paper to prototype to scaled experiments and production, delivering clean, well-tested, well-documented code for multiple teams.