WATERLOO CS · TORONTO

Jack Wang

Computer Science student at Waterloo building browser agents and applied AI.

Available for Winter 2027 co-op in software engineering and applied AI.

Browser Agent Research

WebMCP & Browser Agent Evaluation

Built browser-native tool suites and the evaluation infrastructure used to study how LLM agents act across real web environments.

Problem

Browser agents lost reliability when tool dispatch and element IDs drifted across live pages.

Built

WebMCP tools, a Python evaluation harness, and five WebArena study modes across 9+ sites.

Result

Traced failures across the stack and made the evaluation signal materially cleaner.

9+ sites: tool-enabled environments
10%: avg. success rate increase
40%: token reduction vs. baseline

Read the full case study →

Machine Learning

Chinese ↔ English Transformer

Built an encoder-decoder Transformer in PyTorch with SentencePiece tokenization, evaluation, checkpoints, and beam-search inference.

100K+ entriesprocessed for training

View source ↗

Local AI Automation

WeChat AI Reply Agent

Reads chat screenshots and drafts replies with a vision-language model; manual send is the default, with optional unattended mode.

2 backendsOllama or ChatGPT Web

View source ↗

C++ SYSTEM DESIGN · DEC 2025

Biquadris — Multiplayer Tetris

An X11 game using Observer and Decorator patterns to keep the interface, game state, and runtime abilities decoupled.

Course project · source available on request →

May 2026 — Aug 2026

Huawei Technologies

Co-op Researcher, Web Technology Lab

Markham, ON

Built WebMCP tool suites, a Python evaluation harness, and browser-agent infrastructure for a five-mode WebArena study. Diagnosed tool dispatch and element-ID failures across the stack.

Read the WebMCP case study →

May 2025 — Aug 2025

Innovation AI

Full Stack Engineer

Remote · San Jose, USA

Shipped Vue-ECharts visualization, asynchronous PDF preview, and PostgreSQL-backed store, balance, and transaction-safe cart workflows.

ABOUT

I work across models, interfaces, and evaluation.

Pixel portrait of Jack Wang — Jack, pixel edition

I'm Jack, a Computer Science student at Waterloo. I like finding the seam between model behavior and product behavior, then building the infrastructure that makes both easier to test.

Python
C++
Java
JavaScript
PyTorch
Playwright
Vue.js
Spring Boot
PostgreSQL
Docker