You will evaluate, fine-tune, deploy, and benchmark AI models; design and enforce safety guardrails and sandboxing for agentic systems; safeguard data security and privacy for sensitive research data; and optimize GPU/HPC resource allocation to balance performance, cost, and efficiency as models and tooling rapidly evolve. Experience optimizing inference cost, latency, and resource usage (e.g., KV-cache management, quantization, batching, speculative decoding, high-throughput serving via vLLM/TensorRT-LLM/Triton) and judgment about when a simpler deterministic pipeline is a better fit than an agentic one.