ai focus Link to heading
My current professional focus is agentic data engineering — building production data and governance systems with AI agents as first-class authors, not just assistants.
Thirty years of data warehouse, BI, EPM, and cloud-native analytics underneath. The current emphasis is operating Anthropic Claude Code as a platform — custom skills, scoped permissions, persistent memory, custom MCP wiring — and codifying AI-assisted development as a named methodology (HDD–SDD–BDD) rather than ad-hoc prompting.
Schema-first, governance-only, blast-radius-bounded. Never moves rows, never executes live DDL. The deterministic-vs-non-deterministic line is held where it matters: deterministic for workflows, DDL, and money paths; non-deterministic for synthesis, design, and exploration.
The anchor evidence:
- 10,879 lines of Go in
md-util— SHACL-first governance CLI · v0.8.0 · 2026-06 - 67 codified Claude Code skills for homelab IT operations
- 92 Linear tickets driven end-to-end; 15 closed in one agentic session
- 715 repos / 465,539 commits scanned in 28 seconds by
reposcan - Passed Nate B. Jones 5-Level Prompting assessment (external benchmark)
- First locally trained small model via
LMTRAIN(nanoGPT / MPS)