Hello, I am

Gilbert Wang

Machine Learning Research • Computer Vision • Systems
University of Waterloo — Computer Engineering

Who am I?

An ML-focused engineer who likes clear baselines and reproducible code

I am a Computer Engineering student at the University of Waterloo with hands-on work in time-series (predictive maintenance), computer vision, and tool-driven LLM systems. Recent work includes HSTI (co-authored; STGCN with ~20% lower total loss vs baselines), a photorealistic recoloring pipeline (SAM2 + palette via Qwen3-0.6B on Ollama), and Prompt2CAD (voice/text → parametric FreeCAD via a small RAG flow).

Personal Info

  • Email : g55wang@uwaterloo.ca
  • Phone : +1 (416) 831-3153
  • Location : Waterloo, Ontario, Canada

My Expertise

ML Research & Evaluation

PyTorch, NumPy, TensorBoard; careful baselines, ablations, and reproducible pipelines.


Computer Vision

SAM2 segmentation, OpenCV/NumPy image ops, intrinsic-style recomposition for realism.


LLM Tooling

Ollama (Qwen3-0.6B, Whisper-tiny), lightweight RAG, function planning and orchestration.


My Resume

Experience

2025 — ML Software Intern

Shanghai Jiao Tong University — School of Software

Co-authored HSTI (second author); ran all PyTorch experiments end-to-end (datasets, baselines, ablations). Built small prototypes: Photorealistic Recoloring, Prompt2CAD.


2024 — Software Developer Intern

Measure Tek

Frontend work in Vue; delivered features from architecture through deployment.


2024 — Software Engineering Intern

Schneider Electric

Reusable Java HMI components (e.g., animated conveyor, charts) on Vijeo Designer.

Education

2023 — Present

Bachelor of Computer Engineering

University of Waterloo • Previous term averages: 88% / 89% / 80%.


Before 2023

High School

Focused on programming, physics, and mathematics.

Skills

Python / PyTorch
Computer Vision (OpenCV, SAM2)
LLM Tooling (Ollama, RAG)
C++ / SIMD / OpenMP
Java / Frontend (Vue/React)

88% / 89% / 80%

Recent Term Averages

120

LeetCode Solved

4

Coursera Courses

100%

ECE150 (C++) Grade

My Portfolio

HSTI: multi-sensor predictive maintenance
HSTI

Co-authored (second author). STGCN with ~20% lower total loss vs InceptionTime/DCRNN/TCN. Full PyTorch pipeline, reproducible runs.

Photorealistic Recoloring
Photorealistic Recoloring Pipeline

SAM2 iterative segmentation, palettes via Qwen3-0.6B (Ollama), OpenCV/NumPy recolor + shading recomposition.

Prompt2CAD agent
Prompt2CAD — Voice/Text-Driven Parametric Modeling

Whisper-tiny → Qwen3-0.6B RAG over FreeCAD → ChatGPT to compose code (ops order + parameters).

C++ Matrix Library performance
High-Performance Matrix Library (C++)

Cache-aware blocking, SIMD (AVX2/AVX-512/NEON), OpenMP. Up to 500× vs naïve triple-loop baseline.

Iris Classification Project
Iris Classification (from-scratch NumPy DNN)

Pure NumPy forward/backward passes; >99% on Iris; used for pedagogy and quick prototyping.

2D Shooter Game
2D Shooter (Java)

Small PvP prototype; entity system and input handling for gameplay loops.

STM32 Braille Communication Machine
STM32 Braille Communication Machine

Proof-of-concept device for text communication for deaf-blind users (STM32).

Portfolio Website
Portfolio Website

This site — Bootstrap theme, custom sections, and content system.

Company Website Optimization
Company Website Optimization

UX and performance improvements; content/layout refactor in HTML/CSS/JS.

Send a message

Get in touch

Phone :
+1 (416) 831-3153
Address :
Waterloo, ON
Email :
g55wang@uwaterloo.ca