Case Studies
Not concepts. Not proofs-of-concept. Production systems with real users, real data, and real architecture decisions to defend. Each case study explains the problem, the approach, and what the evidence actually shows.
A personal AI operating layer running on a dedicated Windows machine with 19 MCP tools, DPAPI-encrypted credential vault, multi-AI coordination (Claude, GPT-4o, Gemini), and full audit trails. Not an experiment — a production system for a portfolio of businesses.
Production e-commerce platform (Stripe + Printify POD) rebuilt ground-up: PostgreSQL with row-level security, BullMQ job queue, AI rights screening with 18/18 passing, quality gate, canary traffic controller with 10% increments and atomic rollback. 97 tests. Real users, real orders.
AI-assisted SEO content system for a landscape design business: OpenAI generates 3 posts per day on a schedule, posts enter a review queue with a 3-day hold, a human approves or provides feedback, and the system auto-publishes with image generation and a dynamic sitemap. Human taste as the gate; machine throughput as the engine.
Head of Commercial & IT at a UK cellular pathology company. NHS procurement, HL7-native LIMS operations, digital pathology AI platform evaluation (Indica Labs HALO, Paige AI, Deciphex, 3DHISTECH, Leica Biosystems), ISO 15189-adjacent quality governance, and NGS pipeline context. High-stakes technology decisions with patient-level consequences.
Happy to walk through any of these systems in detail — the technical decisions, the trade-offs, what I'd do differently.