Portfolio · 2026
Zac Matthias
Engineering student at the University of Victoria.
I build software from first principles — and websites for small businesses.
About
I like knowing how everything I make works. When a standard library will do, I won't reach for a package. Where I can build it myself, I will.
That preference isn't ideology — it's leverage. Reliability comes from understanding. Understanding comes from building. The deeper I go into the fundamentals, the fewer surprises there are at the surface.
I'm an undergraduate engineering student at the University of Victoria. I currently work as an Electronics Assembler at TyreSense, hired full-time after a Co-op Production Engineering term where I built the systems below.
Day-to-day I work in C, C++, Python, and JavaScript, with hands-on experience across WebGL, the Web Audio API, Flask, and the PyTorch / HuggingFace stack.
TyreSense
Co-op · Production Engineering. Three internal systems built solo, shipped, and still in daily use. Case studies below.
External company site: Visit TyreSense.com .
TyreSense designs sensor-equipped circuit boards manufactured by an external partner. Across two internal sites — VTS, the R&D and repair lab, and LTS, the production floor I now work on — the shared thread of these projects is the same: turning manual, bottlenecked workflows into systems the people who depend on them can run themselves, with the data trail to back it up.
-
01 / 03
Sensor Tracker
Internal SaaS platform that lets manufacturing partners validate boards independently before shipment. 19,838+ sensors and counting.
Read case study
-
02 / 03
IOTS
Automated testing platform that replaced previously manual test procedures across the production line.
Read case study
-
03 / 03
Production Dashboard
Excel-parsing operational dashboard rendered for TV-mounted monitoring on the floor.
Read case study
Some technical details across these projects remain confidential.
ARES · Language-model research
Three focused experiments in model architecture, training, evaluation, and inference: two transformer studies followed by a character-level recurrent network implemented with standard Python.
Deep dive: Read the ARES case study
ARES is a multi-version academic project covering architecture, training, inference, checkpointing, and model-serving interfaces. Each version is a distinct experiment with a documented objective and set of constraints.
ARES // MODEL 03
-
V1 · Transformer baseline
Custom net & training pipeline
A custom PyTorch transformer experiment and a separate GPT-2 fine-tuning workflow established the training, evaluation, generation, and command-line pipeline.
Source ↗ -
V2 · Modular transformer
Modular refactor
A modular PyTorch codebase with causal self-attention, transformer blocks, validation checkpoints, structured role tags, and configurable sampling.
Source ↗ -
V3 · Standard-library RNN
Character model and HTTP runtime
A one-layer Elman-style character RNN implemented with Python lists, manual backpropagation through time, atomic checkpoints, telemetry, and a browser interface served by the built-in HTTP server.
Source ↗ Engineering reference
Selected Work
A small set, chosen for what they show.
More on GitHub — utilities, scrapers, a browser game, and ASCII video conversion.
Web Design
Custom websites for small businesses.
Alongside my engineering and software work, I independently design and build websites for small businesses in British Columbia.
Projects are built from scratch with responsive design, contact-form setup, performance and accessibility checks, basic search-engine setup, and optional ongoing hosting and maintenance.
I offer both one-time website builds and managed monthly options, with clear pricing and direct support.
Explore small-business web designSkills & Stack
What I reach for
Languages
- Python
- C
- C++
- JavaScript
- HTML
- CSS
- Lua
- Luau
- MicroPython
- R
- MATLAB
Tooling, Workflow & QA
- Git
- GitHub
- VSCode
- Jupyter
- Node.js / npm
- pytest
- Debugging
- Testing & QA
- IIS
- Waitress (WSGI)
- Windows Batch Scripting
Embedded & Hardware
- Arduino Uno R3
- Raspberry Pi Pico W
- Custom PCBs
- IoT
- Embedded Systems
- 3D Printing
- Serial Communication (pyserial)
- Web Serial API
- I2C / SPI Protocols
AI & ML
- Pre-trained Models
- Fine-tuning
- Homegrown Training
- Offline Inference
- Agent Design
Web & Full-Stack
- Frontend Dev (Vanilla HTML/CSS/JS - no frameworks)
- Backend Dev
- REST APIs
- WebSockets (Socket.IO, ws)
- Flask
- Express.js
- SSL / TLS
- CORS & Rate Limiting
- File-Based Storage (JSON, Excel, CSV)
- GitHub Pages / Vercel
- IIS Deployments
Data & Analysis
- pandas
- matplotlib
- openpyxl
- Excel / CSV / JSON Pipelines
- CLI Data Visualization
Engineering Practice
- Systems Thinking
- Real-Time Systems
- Mechatronics
- Prototyping
- Efficient System Design
- Algorithm Design
- Accessibility-First Design
- Local-First Architecture
Contact
Reach out about engineering, software, or a website for your business.
- Email zacmatthiass@gmail.com No, this is not a typo email. I lost my original email account. I know. I should have read the full email before using it. I know it says "ass" at the end.
- GitHub github.com/Chaotic-Mess
- LinkedIn linkedin.com/in/zacmatthias
- Resume Download CV ↗