01 · Assess
AI opportunity audit
I go through your business and find the few places AI is worth the trouble. You get a ranked shortlist, and an honest “skip it” on everything else.
Toronto, Canada · Founder, Olochat AI Inc. · taking on new clients
AI in your company. I'm on it.
A lot of companies are stuck at “we should be using generative AI.” Getting to something real, running in production, is the hard part. That’s what I do.
Trusted with production systems at
- RBC
- BMO
- Bell Media
- CIBC / Simplii
How I work with companies
I work embedded in your team — in your codebase, your standups, your stack — and what I hand over is generative AI running in production, not a deck about it. I handle the strategy up front, the build in the middle, and a clean handover at the end.
02 · Build
Integration & custom builds
RAG assistants, document and email automation, AI features inside your product. I build them to run in your stack and keep running after I leave.
03 · Enable
Team training & handover
I don’t want to be a dependency. I document what I build, walk your team through it, and hand over something they can run and change on their own — with structured AI training through Olochat AI when a whole team needs to come up to speed.
Work with me
I do the work myself, start to finish. Best fit for a focused build or a company’s first AI project.
Work with my team
For bigger jobs, I bring in my team at Olochat AI Inc. — vetted senior engineers who work the same embedded way. You get extra capacity fast, and I stay accountable for what ships.
Security & governance
Enterprise AI lives or dies on the boring parts: where the data sits, who’s allowed to see what, and whether you can prove what the system did. I build for those from the first commit, not the last sprint.
Your data stays in your cloud
Everything deploys into your own AWS, Azure, or GCP account, inside your network boundary. Your data isn’t sent anywhere it shouldn’t be, and it’s never used to train a model.
Access control at the retrieval layer
Per-user permissions are enforced before anything reaches a model, not bolted on afterward. People only see answers built from documents they’re already allowed to read.
Auditable by design
Every answer cites the sources it came from, and key decisions are logged. When someone asks why the system said what it said, you can actually show them.
Model-neutral
Claude, OpenAI, Azure OpenAI, or open-weight models on your own hardware — chosen to fit your risk, latency, and cost profile. You’re never locked to one vendor.
Inside your process, not around it
I work under your NDA and MSA, through your security review, and alongside your InfoSec and platform teams — not in a shadow project that has to be re-approved later.
Evals and monitoring from day one
Quality is measured against a real evaluation set before anything ships, and watched for drift after. “It feels better” never gets to decide a release.
Selected work
Olotalk
Shipped a production RAG support agent for Canadian SMBs — the flagship product of my company, Olochat AI Inc. A live-chat assistant that answers from a company’s own site content: DOM-aware chunking and Claude-based enrichment over Qdrant, PostgreSQL, Redis, and ClickHouse, deployed on Kubernetes with zero-downtime rollouts.
Cloud Migration Diff
Made cloud-migration config reviews readable. A Next.js and Lambda tool that parses and color-codes the differences between on-prem and cloud JSON configs, with secure access boundaries via AWS IAM roles, policies, and STS.
Risk Explorer
Let risk analysts trace relationships across their data. An interactive tool of draggable, interconnected charts wired together with Dijkstra and topological sorting, plus exploratory LLM tooling for data insights.
Experience
Nearly 20 years building production systems, from telecom and banking platforms to AI. Trilingual (English, French, German), based in Toronto. Recent roles:
2025 – Now
Founder & Principal
AI training for teams, and AI products of our own — including Olotalk, a production RAG chat and support platform I architected and shipped end to end.
2009 – Now
Principal Developer & AI Consultant
Independent practice · Toronto
Full-stack and AI/LLM delivery for fintech, media, and public-sector clients, end to end.
2024 – 2025
Senior Full-Stack Developer
BMO · Cloud Center of Excellence
Built cloud-migration tooling on Next.js, Node.js, and AWS Lambda.
2022 – 2024
Senior Full-Stack Developer
RBC · Enterprise Risk Management
Built an interactive risk-visualization tool with graph algorithms and LLM-assisted insights.
2018 – 2020
Senior Full-Stack Developer
MobileLIVE
Led fido.ca’s AngularJS to Angular 7 migration; built CIBC/Simplii money-transfer features.
2018
Senior Frontend Developer
Bell Media
Built features for the Vidi video player used across 25 Bell brands.
Writing
The gap between an AI demo and production 2026 · 07 RAG in production: what the demos don’t tell you 2026 · 02New posts land here first. No newsletter, no popups. Just a page that updates.