Matthew Li

I'm currently studying computer science at the University of Waterloo, and have a strong interest in building impactful and challenging software.

A couple of transit related projects have recently been in the news, including CityNews, BlogTO, and Waterloo News.

In my free time, I enjoy reading, skateboarding, and playing basketball, which are often interests I incorporate into my projects.

Work

Mercor
Software Engineer — RL environments and post training
Hecaton
Software Engineer — building a React Native app for a Canadian crypto exchange
FriedmannAI
Member of Technical Staff — financial portfolio management/forecasting + applied ai
Waterloo Data Science Club
Machine Learning Engineer — guardrail research + response, presented research at CUCAI
Events Coordinator — planning events and things of that nature
DevFortress
Software Engineer — developed ML training libraries + quality assurance for Shopify apps

Some projects

  1. i.
    sidewalk-shed — route planner through NYC under maximum scaffolding coverage (5k+ users)
    Maps a walking route through New York City under as much scaffolding as possible. Useful if it's raining and you forgot your umbrella or something.
  2. ii.
    cramrapp — group study session and progress tracking, competitive features coming (1k+ users, in progress)
    Study tracking platform for groups — log sessions, measure progress, and compete with friends. Competitive leaderboard features actively in development.
  3. iii.
    ttcleaderboard — live ranking of TTC streetcars by speed (200k+ users)
    Real-time leaderboard tracking the speed of every Toronto streetcar. Featured in CityNews, BlogTO, and Waterloo News.
  4. iv.
    protosynthesis — drag and drop workflow builder with 30+ templates, webhooks, and MCP connections (in progress)
    Visual workflow automation builder. 30+ prebuilt templates with support for webhooks and Model Context Protocol (MCP) connections. Actively in development.
  5. v.
    url shortener — production-grade url shortener with full observability stack
    Built with Nginx, Redis, Prometheus, Grafana, Loki, and Fail2Ban. Full metrics, logging, and rate limiting — proper infra, not a toy.
  6. vi.
    everything ever — WebGL map of the startup/open-source landscape that finds genuinely unoccupied idea space
    Self-contained Python + WebGL system that maps 6,000+ YC companies and 1,400+ GitHub projects, uses novelty search to find genuinely unoccupied regions, has an LLM write concrete product ideas into those gaps, then audits whether each idea actually landed in empty space. Embedder, SVD, graph layout, and renderer all hand-written — the only dependencies are Python and numpy.
  7. vii.
    apply-watcher — production job-monitoring infrastructure built for zero2sudo, an early-tech-career creator with 93K+ Instagram followers and a 250K+ community
    • Engineered a lightweight HTTP monitoring service for Google and other early-career roles, polling job pages every 5 minutes and detecting server-rendered application availability without browser automation
    • Built 3 production deployment variants: zero-dependency Python 3 with native macOS notifications; Node.js 22 AWS Lambda with EventBridge scheduling, SSM Parameter Store, and iMessage delivery; and a configurable Python notifier supporting Slack, email, and webhooks
    • Designed for 24/7 unattended operation, with resilient HTTP fetching, persistent state, idempotent change detection, and duplicate-alert prevention across 1,000+ tracked listings
    • Eliminated browser automation to minimize compute and operational overhead, enabling low-latency, continuous polling across multiple notification and deployment environments