I left everything comfortable to build a life of my own. Turns out the difficulty was the point. I've been looking for the next one ever since.
Off the clock
I fly standby and debug production. Both run on optimism.
Swimming
Coding
Football
Reading
Board games
Music
What I do
Most systems are quietly not doing what everyone thinks they're doing. I find that more interesting than almost anything else.
Nothing was failing
New enrolments had stopped reaching the fraud screening service. Every dashboard was green, because nothing had failed — a thing that was supposed to happen simply wasn’t happening.
Finding where they went took a trace across two AWS accounts and two regions, in Athena. An upstream request validator was rejecting any request with a null middle name — exactly what the web form sends when someone skips the field. A 400, before the fraud call ever happened.
Downstream, error handling read any fraud-service error as a pass. So enrolments succeeded. Quietly, and with no fraud case ever created.
Tens of thousands of profiles, over about two and a half weeks. The fix belonged to another team. The work was noticing, and then proving it.
Most of my job is finding out something has been broken since 2023 and nobody noticed.
Partner linking, end to end
Linking and delinking between Aeroplan and hotel, retail, and mobility partners.
Spec corrections where a partner's documented hostname didn't exist. HTTP status semantics, including a 204 No Content response that carried a body — which violates the spec and made every success look like a server error one layer up. And a full triage of one day's failures, categorized by root cause and handed to the upstream team.
Consent platform migration
Consent and preference management moved from one vendor platform to another with no downtime for marketing operations.
A parallel ingestion path ran both systems side by side through cutover; the old one was retired in stages. “No downtime” meant no gap in consent state — a legal requirement, not a nice-to-have.
Is it a lot?
- requests a day to a core member-context service in production
- 2 million+
- of members in the loyalty program
- Tens of millions
- of reconciliation events a day
- Hundreds of thousands
At this scale, 'rare' means daily.
Am I good?
Systems I’ve integrated with
I built these from the Air Canada side.
Where I’ve worked
2024 –
2023 – 24
2021 – 22
The paper
MSc Applied Computer Science
Montreal · 2022 – 24
Certified, six times over
AWS Developer – Associate- AWS AI Practitioner
- AWS Cloud Practitioner
Azure Developer Associate- Azure Data Fundamentals
- Azure Cloud Fundamentals
What I build with
- Languages & APIs
- Platform
- AWS
- Lambda, DynamoDB, API Gateway, MSK/Kafka, Glue, S3, Athena, EventBridge, CDK, CloudFormation, CloudWatch, X-Ray
I leave systems more explainable than I found them.
That’s the tour.
Built mostly by an AI, in about a week. I'm curious where that goes.