LLM Optimization Engineer

Responsibilities

• Design and implement efficient parallel computing strategies and memory management mechanisms to improve end-to-end throughput and latency
• Develop and optimize high-performance training and inference frameworks, maximizing hardware compute and memory bandwidth utilization

Qualifications

• Proficiency in Python and C++, with strong foundations in data structures, algorithms, and systems programming
• Solid experience with PyTorch, including a deep understanding of model execution workflows, operator invocation, and computation graph mechanisms
• Familiarity with high-performance computing (HPC) concepts such as parallel computing, memory hierarchy, and operator fusion
• Basic understanding of accelerator architectures (e.g., GPU, NPU), including compute units, memory systems, and communication mechanisms

Preferred Qualifications

• Experience with mainstream LLM inference acceleration frameworks such as vLLM and SGLang, with hands-on performance optimization experience
• Familiarity with techniques such as KV cache optimization, attention optimization, operator fusion, and low-precision computation (e.g., FP8, FP4)
• Experience in productionizing large model training or inference systems, with end-to-end performance optimization experience