Consulting

I build LLM agents for regulated environments: Canadian banks, the Government of Canada, telecom, and wealth management. Fourteen years of full-stack work in those places; LLMs since 2022. In this world, audit trails, data residency, PII redaction, human-in-the-loop approvals, and prompt and version governance are not optional.

Most teams building LLM agents optimize for the demo. The constraints that actually exist in a bank or a government department (what data the model can see, what it can log, who must approve a decision before it ships, how you prove what the system did six months later) get bolted on late, expensively, and often badly. I build the other way around: the constraints are the architecture.

What I do

• Strategy and architecture for agentic systems that have to survive audit and regulator questions.
• Hands-on implementation of guarded, observable LLM agents: guardrails, redaction, audit trails, evals, and human-in-the-loop checkpoints.
• Standalone component builds: I build a component independently and hand it over for your team to integrate, no embedding required.
• Innovation work: experimental builds and proofs of concept in environments that don't tolerate surprises.
• Team upskilling on prompt and context engineering with compliance built in, not bolted on.

How I engage

Advisory, embedded build, standalone build, or training, usually some combination. Short engagements are fine.

Contact

Reach me on LinkedIn. A few sentences about your team and what you're trying to build is enough to start.