We are looking for a systems-minded engineer who lives at the intersection of large-scale model inference, distributed systems, and performance optimization. This role focuses on post-training and inference infrastructure, with particular emphasis on P/D disaggregation, KV cache lifecycle management, and efficient offloading mechanisms across both inference and reinforcement learning (RL) systems.

Research and deeply understand modern LLM inference frameworks, including: Architecture and design tradeoffs of P/D (prefill / decode) disaggregation; KV cache lifecycle, memory layout, eviction strategies, and reuse; KV cache offloading mechanisms across GPU, CPU, and storage backends; Analyze and compare inference execution paths to identify: Performance bottlenecks (latency, throughput, memory pressure); Inefficiencies in scheduling, cache management, and resource utilization; Develop and implement infrastructure-level features to: Improve inference latency, throughput, and memory efficiency; Optimize KV cache management and offloading strategies; Enhance scalability across multi-GPU and multi-node deployments; Apply the same research-driven approach to RL frameworks: Study post-training and RL systems (e.g., policy rollout, inference-heavy loops); Debug performance and correctness issues in distributed RL pipelines; Optimize inference, rollout efficiency, and memory usage during training; Collaborate with research and applied ML teams to: Translate model-level requirements into infrastructure capabilities; Validate performance gains with benchmarks and real workloads; Document findings, architectural insights, and best practices to guide future system design