Talks & Speaking

Practitioner, not pundit.

I speak about applied AI in regulated, high-stakes environments — field reports from production systems, not slideware. Conference sessions, technical meetups, and executive briefings.

What I speak about

Pharma / Life Sciences

AI in Regulated QA: What Actually Works, What Breaks

Most AI demos in regulated industries fall apart the moment a real QA reviewer touches them — hallucinated citations, no audit trail, no way to trace a decision back to source. A practitioner’s field report from building a production RAG-based QA copilot for pharma cold chain: the failure modes that killed the first three prototypes, the human-in-the-loop pattern that finally worked, and where AI belongs in compliance-critical workflows — and where it absolutely doesn’t.

45-min session or 60-min workshop

AI / ML Engineering

Human-in-the-Loop Isn’t a Safety Net: Designing It as Primary Architecture

Most teams bolt human-in-the-loop onto an AI system as a backstop for low model confidence. That’s the wrong frame. In high-stakes domains, HITL is the primary architecture and the AI is the assistant. A walk through a production system where the human is the decision-maker by design: the UX patterns that scale, the ones that turn reviewers into rubber-stampers, and audit trails that capture reasoning instead of just outputs.

30–40 min technical session

Executive / Leadership

First-Order AI: Why Most Enterprise AI Initiatives Plateau Before They Pay Off

Most organizations are doing first-order AI: bolt a model onto an existing workflow, declare victory, watch adoption stall. Second-order is product-aligned. Third-order is the whole picture — engineering org, risk posture, cost model, team structure. The Three Orders framework with examples from regulated industries, where staying at first-order isn’t just stalled ROI — it’s compliance debt. Leaders leave with a 90-day diagnostic for their own AI portfolio.

30-min strategic talk, panel-friendly

Recent talks

AI Tinkerers Boston — GTM / Agentic AI Launch June 2026

I Built An Agent To Be Me On LinkedIn (And It Failed Three Times)

A live demo of a multi-source, voice-conditioned content pipeline: daily tinkering capture, GitHub scans, and web research aggregated weekly into drafts written in my voice — with a human editorial gate, and the three failure modes it took to get there.

AI Tinkerers Boston — Boston Tech Week May 2026

The Duct Tape Was Hiding Three Bugs: An Agent Migration Story

A four-phase Claude tool-use agent that turns a home-project description into a sourced project blueprint — and the three bugs that only became visible the day the JSON-parsing duct tape came off in favor of structured outputs.

Booking a speaker, or running a CFP? Also available for private training and workshops for engineering teams and leadership.

hello@thirdorderlabs.io