About
I build the platforms that make AI dependable.
I’m Mark Thebault, a principal architect working across Agentic AI and platform engineering. For most of my career I’ve sat where ambitious ideas meet the realities of running software at scale — translating what a team wants to build into systems that hold up when real users and real traffic arrive.
Lately that means agentic AI: orchestration, tool use, evaluation harnesses, and the infrastructure that lets teams ship agents they can actually trust. I care about the unglamorous parts — observability, cost, failure modes, developer experience — because that’s where the difference between a demo and a product lives.
This site is where I think out loud: essays, field notes, and the occasional opinion, mostly aimed at the engineers and architects doing this work alongside me.
How I work
- 01
The platform is the product
Agents are only as reliable as the system beneath them. I design the orchestration, evaluation, and guardrails as first-class infrastructure — not afterthoughts bolted onto a prototype.
- 02
Production is the real benchmark
A demo proves something is possible. Latency budgets, failure modes, cost ceilings, and on-call pages prove it’s real. I optimise for the second list.
- 03
Leverage teams, not just code
The highest-leverage architecture work is the kind that makes a whole team faster: paved roads, sharp abstractions, and decisions that age well.