How I Built a Zero-Trust LLM Security Proxy

A few weeks ago, I found myself staring at a problem that kept getting worse. Every team in the org was spinning up their own LLM integrations — some through OpenAI, a few through Ollama, one team eve