Field Notes · Production AI

Writing on production AI
in regulated industries.

Long-form essays, the bi-weekly Field Notes newsletter, and dispatches from the inside of healthcare, energy, and finance AI deployments. No benchmarks. No demos. What actually breaks in production — and how I fixed it.

Bi-weekly · #HumanWritten · No spam · Unsubscribe anytime
Long-form essays

Medium archive

Agent Security LLM Observability Regulated AI MLOps Healthcare RCM Genomics Production GenAI Developer Tools
Claude Code Authentication: Cross-Platform Setup ScriptNew
The Five-Layer Defense: How Bulwark Prevents AI Agent Hijacking in ProductionNew
The Web Is Now Weaponized Against Your AI AgentsNew
I Built What Clinical Genomics Has Needed for 20 Years. It Goes Live Today.
Built 11 Autonomous Agents to Fix Healthcare Revenue Cycle
78% of PyTorch Models Never Reach Production. I Built the Fix.
The Invisible Reason 70% of Enterprise AI Programs Fail — And the Tool I Built to Fix It
Your LLM App Is Degrading Right Now. You Just Cannot See It.
The LLM Metrics That Actually Matter in Production
Building a Production-Grade HIPAA-Compliant Healthcare Q&A Platform with Claude AI and Codex
Why Enterprises Waste 40% of Their AI Budgets (And the Open-Source Solution I Built)
Full archive on Medium →
Bi-weekly newsletter

Field Notes:
Production AI

Bi-weekly dispatches on what is actually breaking in production AI. Each issue covers one specific failure mode I have seen — from RAG groundedness collapse, to agentic eval drift, to compliance-breaking audit gaps — and the engineering decisions that surfaced or solved it. No benchmarks. No prompt-engineering tips. Field-grade only.

Why Agentic AI Changes the Evaluation Problem Entirely
The $4.1B Acquisition Lesson Nobody Talks About
Browse the full Field Notes archive on LinkedIn
Coming Next

In the queue

The Rule of Two Doesn't Survive Contact With Healthcare
Regulated AI Incident Review — Issue #1