Your teams are already using AI tools — ChatGPT, Copilot, Claude — without anyone having defined what can be sent to them and what cannot. Or you want to deploy agents capable of automating business processes, without knowing how to isolate them from your critical systems.

I work on two areas: advice and usage policy (data classification, guardrails for cloud AI services) and on-premise deployment (open source models on your infrastructure — Ollama, vLLM, LiteLLM — for data that cannot leave your perimeter). In both cases, I design agents with the required isolation mechanisms.