Economics, Carleton University
BA in Economics. I learned to ask what a number is actually measuring.
Character profile: Bryan Chuinkam
ブライアン・チュインカム
- Class
- Senior Full-Stack & AI Engineer
- Base
- Ottawa, Canada
- Experience
- 7+ years shipping production software
- Languages
- English, French
- Specialty
- AI products with the model on a short leash: narrow jobs, checked outputs, humans approving what matters
- Off duty
- Father and husband. Sports fan. Competitive about everything, including daily trivia.
- Weakness
- Can't leave a flaky test alone
The real one, for comparison
Yes, the glasses are real. The speed lines follow me everywhere.
I started in economics, fell for data, and spent the last seven years shipping production software for other people's teams. Now I build my own products end to end: problem, prototype, pilot, CI/CD, and the support inbox. I like work where getting it right matters, like money, kids' learning and personal data.
Backstory arc
- 2013
- 2017
Business Intelligence Systems, Algonquin
Postgraduate certificate. Data got its hooks in.
- 2017
Insight Analyst, MD Financial Management
Built database applications on SQL Server for finance, marketing and research teams, working inside each team's workflow.
- 2019
Software Developer, R&D, WiseWithData
Built SearchParty, which cut client onboarding effort by 80%+. Shipped a production RAG assistant over the company codebase. Moved client ETL from SAS to PySpark with up to 5x throughput.
- 2023
Full-Stack Web Development, Lighthouse Labs
Diploma. Added the front half of full stack.
- 2026
Cloud Development & Operations, Algonquin
Postgraduate certificate. Capstone: a multi-tenant RAG-pipeline SaaS on Azure.
- Now
Four products in the field
OwlReader, Pinalty, ConveneSpace and officeHours, each live or in pilot.
Techniques
- LLM integration
- Anthropic Claude API, OpenAI via Vercel AI Gateway, Vertex AI / Gemini, Azure AI
- Structured outputs
- Forced tools, JSON-schema validation, and code that checks before it trusts
- RAG pipelines
- PGVector, SparkNLP, Ollama, multi-step orchestration
- Postgres that guards itself
- Row-Level Security, permission-checked functions, pgTAP tests
- Full stack
- TypeScript, React, Next.js, Python, FastAPI, Django
- Data at scale
- Apache PySpark, Spark SQL, SQL Server
- Cloud & delivery
- Azure, AWS, GCP, Vercel, Docker, GitHub Actions, Azure DevOps
- Testing
- Vitest, Jest, PyTest, Playwright, TDD
The code
Rules every product here follows.
A human approves.
The model drafts. A person decides anything that matters.
Keep only what you need.
No raw model responses stored, and no data I don't need collected.
Keep it in Canada.
Canadian data residency whenever users are Canadian.
Test the locks.
Row-Level Security is only real if a test tries to break it.
Side quests
Teaching is how I check that I actually understand something.
- Taught a hands-on PySpark data class for underrepresented high school students, with Colourfully Digital
- Volunteer instructor with Black Boys Code Ottawa
- Trained client teams at WiseWithData and spoke at a Python user group
- Writes about AI-assisted development on Medium
- Read my writing on Medium
Seen enough? Let's talk.
Read the afterword