stephen

ontario, canada

eng @ uw

programmer, ml/ai enthusiast, philosophy larper

"truth is found neither in the thesis nor in the antithesis, but in an emergent synthesis which reconciles the two."

~ georg wilhelm friedrich hegel

projects

maze env

[github]

python, numpy, matplotlib, pandas, reinforcement learning

benchmarking temporal difference control algorithms on grid-world maze navigation to analyze convergence, variance, and behavioral differences.

infralens

[github]

digitalocean, snowflake, fastapi, nextjs, terraform, elevenlabs, twilio

an automated cloud rightsizing pipeline using metrics, terraform, and llms. won the mlh best use of digitalocean award at conhacks 2026.

city mind

[github]

react, fastapi, python, sqlite, langgraph, claude sonnet, pytest

a governed municipal data gateway enforcing role-based access control and privacy. won 1st place for the city data os problem statement at fci x langchain hackathon.

mnist classifier

[github]

numpy, python, matplotlib

developed a fully-connected feedforward neural network utilizing only numpy without high-level ml libraries, achieving ~95.8% training accuracy.

pkdex

[github]

react native, pokeapi, langchain/langgraph

mobile application that provides detail pokemon information and battle analysis.

recent posts

building infralens

how my team built an AI-powered Terraform optimization agent and won the MLH best use of DigitalOcean award at ConHacks 2026

building city mind

a description of the solution i built during the FCI x LangChain hackathon

presuppositions

hello world