Years of Experience 1+ Years Demonstrated experience deploying large open-weight LLMs on GPU infrastructure on-premise (multi-GPU, H100-class nodes). Inference serving and optimization: vLLM / TensorRT-LLM or equivalent, quantization, batching, and sizing for concurrent users. Building LLM-backed features on top: assisted review, natural- language querying over structured data, and bulk triage workflows. Comfortable working in an air-gapped / restricted-network environment with no external API dependencies.