A
Book

[04] · About

One engineer.
A team’s worth of work.

I’m Anzhe Dong (Barton in English) — a Toronto-based AI engineer building production agentic systems end-to-end. I was the third hire and founding AI engineer at TripFix, where I took an empty repo to a multi-agent product running thousands of claims with deterministic citations and five-evaluator quality gates.

Before AI, I spent three and a half years at CGI doing serious data engineering on Spark, Airflow, and Databricks — the kind of work where reliability and observability matter as much as the algorithm. That instinct shows up in everything I build now.

[01]Career timeline

  1. 2025 — 2026

    Settled · TripFix

    Role

    Founding AI Engineer · 3rd hire

    What I did

    Took the AI surface from blank repo to a multi-agent claim co-pilot used in production. Owned multi-agent orchestration, eval harnesses, vision reasoning, MCP-style skills, and the full Python + Laravel + Postgres stack.

  2. 2022 — 2025

    CGI

    Role

    Senior Consultant · Data Engineering

    What I did

    Spark, Airflow, Databricks at enterprise scale. Shipped pipelines that moved hundreds of millions of records a day for finance and telecom clients.

  3. 2021

    Stanza

    Role

    Software Engineering Intern

    What I did

    First taste of production code in a small product team. Got hooked.

  4. 2017 — 2021

    University of Toronto

    Role

    BSc Computer Science

    What I did

    Distinction. Specialised in AI and software engineering streams.

[02]How I work
01

Ship, then prove it.

I’d rather have an agent in prod with a five-evaluator harness than a perfect agent in a notebook. Evaluation in production is louder than evaluation in theory.

02

Tight loop > perfect plan.

I’d rather show you something running on Friday than a polished doc on Monday. Plans bend the moment the model sees real data.

03

Tooling as leverage.

Every engineer I admire has built quiet, personal infrastructure that lets them do a team’s worth of work. I optimise for that compound.

04

Plain English first.

If I can’t explain the architecture to a non-technical operator, the design is wrong — not their understanding.

[03]Off the record
Lives in
Toronto, Canada
Languages
English · 中文 (Mandarin)
Coffee
Single-origin Ethiopian
Reading
Patrick Collison’s blog rotation
Favourite model
Sonnet for thinking; Opus for coding
Calendar handle
barton97daz@gmail.com

[04] · Talk

Curious whether
we’d work well together?

Coffee chats, career advice, mentorship, role conversations — book a slot below or just send a note.