Design and implement core, high-performance services for an enterprise-scale AI policy engine. Own architecture and APIs, make tradeoffs across correctness, consistency, availability and latency, and work across stack from data models to product surface to ensure auditable, reliable governance under concurrent load.

Build production systems that probe and modify LLM internals to improve reliability and enforce deterministic policy. Implement mechanistic-interpretability techniques (activation patching, control vectors), work with weights/activations, build evaluation and deployment loops, and design feature-level intervention systems for enterprise inference.