Senior LLM Tooling and Interpretability Engineer

This profile highlights extensive experience in data science and analytics (~15 years), with a recent and strong focus on LLM internals, tooling, and interpretability. The individual has been actively working on LLM introspection tooling and observability, specifically around token-level behavior during inference and experimenting with inference-time interventions. A notable achievement is the development of a demo tool (Cartogemma on Hugging Face Spaces) that exposes per-head projections, token rank changes, top-k next-token branches, and capabilities to mute heads, inject tokens, or rewind context. The background includes translating technical work into decisions and policy for leadership, and a foundation in philosophy, applied linguistics, and NLP, including teaching a course on propaganda. The individual is seeking roles in LLM tooling, evaluations, interpretability, or applied AI where understanding model behavior is crucial. They are proficient in Python, Rust, SQL, embeddings, and local LLM infrastructure. Preferred work arrangement is hybrid or remote, and they are not willing to relocate.