No estimates, no promises. Each card links to measured production results with full methodology.
Enterprise deployment on hardened infrastructure went sideways. Locked out of production VMs 4,000 miles from the hardware.
AI governance hooks firing after the write don't prevent violations. Three rounds of agentic testing proved only PreToolUse hooks actually block.
Production use of Claude Code surfaced real bugs and missing features in the tool itself. Filed issues directly on the Anthropic repository.
Assumed backup coverage existed. AI-assisted audit of Azure SQL, Managed Instances, Cosmos DB, and Storage revealed zero verified protection.
100+ daily API alerts with no triage. Built an AI-orchestrated pipeline that transforms noise into actionable intelligence.
Azure's naming constraints conflict: length limits, uniqueness requirements, and human readability. Solved with a deterministic algorithm.