// full-stack & distributed systems

MatthewHahn

$./career --company="Amazon" --role="tech lead" --location="Toronto"

Full-stack engineer at Amazon in Toronto. I lead the Amazon Author page across worldwide marketplaces, and before that built inbound placement systems for Fulfillment by Amazon. On the side, I ship my own products, most recently DateCraft.

6+
years at Amazon
8+
years building software
12
marketplaces
see my work

./about

Queen's University
BASc Computer Engineering
Class of 2017

Matthew Hahn

// me

I like understanding how things work. I apply the engineering method to everything I build, from the design document to the code that ships.

I started my career at Grip Limited, a Toronto ad agency, where I worked across the stack to rebuild their ad platform as a serverless product on AWS. In 2020 I joined Amazon, and I've spent the 6+ years since on systems that power its worldwide marketplaces. Today I'm the tech lead for the Amazon Author page, where I work on how readers discover books. Working at that scale has taught me to care about reliability, operational readiness, and code nobody has to debug at 3am.

Outside of work, I build my own products: DateCraft is live, and Beemoot is in beta. I also write here about taking things from zero. I've been a drummer my whole life, playing in jazz, brass, and musical pit bands, and I was part of Queen's Music Club through university.

// recognition

  • Advisory BoardQueen's Electrical and Computer Engineering Corporate Advisory Board, current member
  • 1st PlaceQueen's IEEE Student Papers Competition, 2017
  • FinalistProfessional Engineers Ontario Kingston Chapter Student Papers Competition, 2017
  • Dean's ScholarQueen's University Faculty of Engineering and Applied Science, 2017

./work

Work Experience

Amazon

current
2020 – Present
Toronto, ON

  1. Software Development Engineer IITech Lead, Amazon Author Page (Books)

    Sep 2025 – Present
    • →Tech lead for the Amazon Author page across Amazon's worldwide marketplaces. Leading its migration from a legacy framework to a modern component architecture; shipped the first component and wrote the architecture spec the team adopted.
    • →Diagnosed a 15% click-through regression on the Author page Follow button after migration. Traced it to component-lifecycle differences between the old and new frameworks and shipped the fix.
    • →Wrote the engineering strategy and four workstream specs for AI-accelerated feature delivery on Author Central, a brownfield system of 8+ legacy services. Reviewed by 8 engineers and adopted as the team's approach.
    • →Designed the strategy's measurement layer: one schema-validated record per change. Each record captures cycle time, rework, and escaped defects, pairs every speed metric with a quality metric, and holds an estimate committed to git before work starts.
    • →Built the Author team's first LLM agent, orchestrating several MCP servers. Benchmarked against the baseline assistant, it answered twice as fast with fewer tool calls and better domain recognition. Introduced a persistent-context pattern that cut AI session cold start from 5–15 minutes to near zero.
    JavaTypeScriptAWSCDKLLM AgentsMCP

  2. Software Development Engineer IIFBA Inbound Placement

    Dec 2022 – Sep 2025
    • →Designed a new routing path that sends FBA inbound inventory through sort centres; launched in the US and scaling across the fulfillment network.
    • →Reduced placement-fee reporting latency from 24 hours to 4 hours.
    JavaAWSDistributed SystemsData Pipelines

  3. Software Development Engineer I

    Mar 2020 – 2022
    • →Joined as SDE I; promoted to SDE II in about 2.5 years.

Grip Limited

2017 – 2019
Software Engineer·Toronto, ON

Rebuilt Grip's proprietary ad software from the ground up into a fully serverless, scalable architecture on AWS.

  • →Rebuilt the ad builder as a serverless platform on AWS: 30+ Python REST endpoints on API Gateway and Lambda over a DynamoDB schema I designed.
  • →Implemented OAuth 2.0 with Cognito, and real-time chat over WebSockets using AWS IoT Core as the message broker.
  • →Automated CI/CD with Bash, Git, and Docker.
PythonTypeScriptReactAWS LambdaDynamoDB

./shipping

What I'm Building

Live

DateCraft

Plan dates worth remembering.

Tell it who you're seeing, the mood, and when, and it builds the whole evening: dinner, the thing after, and the nightcap, matched to what you both like and all within a short walk. Live with a free tier and a paid Plus plan.

// under the hood Builds multi-stop plans from live venue data, spacing stops by walking distance and checking them against the forecast. Recurring plans are generated ahead of each date night, and post-date feedback shapes future suggestions.

ReactTypeScriptSupabase
datecraft.ca
In beta

Beemoot

Find your next best friend. Get involved in your community.

A social platform for discovering local communities and the people in them. Built in React and TypeScript.

ReactTypeScriptTailwind CSS
beemoot.com
In development

Vault

A single place to store all of your favourite things, and share with friends.

A personal catalogue of the things you care about, with a record of how long you've loved each one.

ReactTypeScriptTailwind CSS
viewmyvault.app

./projects

Engineering Projects

Selected university work, 2014–2017.

2017

Drone Detection System

Award-winning final year engineering project. C++ program that analyzes radio signals to detect drones operating nearby, capable of detection up to the maximum physical range of the drone controller.

Wrote C++ code for signal data processing, MATLAB scripts for signal analysis, and Python scripts for raw data parsing. Won 1st place at the Queen's IEEE Student Papers Competition.

C++PythonMATLABSignal Processing
2016

Garnet Language Compiler

Built a complete compiler for Garnet, a new programming language based on Ruby. Implemented all phases: scanner, parser, semantic analysis, and code generation in S/SL and PT Pascal.

S/SL is the same technique used in IBM's COBOL compilers. A strong exercise in the fundamentals of software engineering.

S/SLPT PascalCompiler Design
2016

Game AI with Neural Networks

Neural network that reads in-game player actions and predicts engagement, frustration, and challenge levels, trained on a dataset of Mario players.

Tested multilayer perceptron with backpropagation using 1–5 hidden layers with varied inputs in MATLAB. Outputs help game developers understand user demographics and adjust gameplay accordingly.

MATLABNeural NetworksMachine Learning

// more from university

  • 2016Queen's Rocketry WebsiteDesigned and built the website for the Queen's Rocketry Engineering Design Team.
  • 2015QBNB: House Renting ServiceAirbnb-style rental platform for Queen's alumni, built with two teammates.
  • 2015Arduino Robotic ArmArduino-controlled robotic arm that moves and grabs objects in three dimensions using servos.
  • 2015Hangman++Hangman game with three difficulty levels and custom game modes, built by a team of six.
  • 2014Solar-Powered Digital SignPortable, programmable, solar-powered sign, donated to a local organization.

./blog

Writing

// no posts yet

First post coming soon.