🔬 Six Sigma MBB engagements — DMAIC, FMEA, SIPOC, CTQ flowdown, control plans. Targets are statistically derived and rolled up through process capability, with Z ≥ 6.0 as the binding metric — not time-based proxies.

🤖 AI agent architecture — RAG, eval-first LLM systems, agentic OS for solo B2B consulting. Production-grade: gates in code, not prompts. The flagship OS has 8 pillars, 1,300+ tests, deterministic boundaries.

🏭 OT + industrial cyber — Process-defect framing for cyber-induced downtime. Five exposure classes, 12-mode FMEA, Poka-Yoke enforcers, AI-augmented Layer 3. CISO is executive sponsor, sole signatory.

Frameworks — DMAIC · DMADV · FMEA · SIPOC · CTQ Tree · Poka-Yoke · SPC · Kaizen · OEE · COPQ

Engineering — Python · TypeScript · React · FastAPI · LangChain · ChromaDB · Ollama · Playwright

Operating principles — Eval-first agent builds · test-driven development · capability-index reporting · data-blind architecture · CISO kill-switch on all autonomous agents

"Targets must be statistically derived, defect-based, and rolled up through the hierarchy of process capabilities — not selected as arbitrary time thresholds."

• The capability index is the binding metric. Time values are operational translations of Cpk and Z, never the contract themselves.
• Gates in code, not prompts. Every AI agent is bounded by deterministic checks; the LLM never self-policies safety.
• Anonymised by design. The data-blind trust boundary lets the engagement run without exposing customer internals.
• CISO is the executive sponsor, sole signatory. I run the methodology; the sponsor owns the risk appetite and the tollgates.