Anshi Wang
anshiwang16@gmail.com (647) 512-6757 Richmond Hill, ON linkedin.com/in/an-shi-wang github.com/anshiwang8
Education
McMaster University
Hamilton, ON
Expected 2031
Bayview Secondary School
Richmond Hill, ON
June 2026
- Graduated with a 94.3% average as an Ontario Scholar and four-time Honour Roll recipient
- Named Athlete of the Year; OFSAA Badminton Bronze Medallist; Vice President of the Athletic Council and the Bayview FBLC
Experience
Instructor & Center Assistant
Oct 2025 to Jun 2026
Richmond Hill, ON
- Delivered personalized math and reading instruction to 30+ students, adapting to each learner’s pace and level
- Owned grading and progress tracking across the center, ensuring accurate performance monitoring
- Coordinated with staff and parents to maintain an organized, focused learning environment
Front & Back of House
Apr 2025 to Oct 2025
Richmond Hill, ON
- Kept service running smoothly in a high-volume restaurant, moving fluidly between roles under pressure
Projects
Itinerary | Next.js, TypeScript, Llama 3.3 70B, Google Maps Platform, Redis, Playwright
2026 to Present- Founded and shipped a live AI day-planner (itinerary-six.vercel.app) that turns one sentence into a time-sequenced day out: real venues, travel legs between them, and per-stop timing
- Architected a pipeline that confines the language model (Groq Llama 3.3 70B) to semantic parsing and venue selection, while deterministic code enforces every verifiable fact: hours, price, rating, weather, and routing via the Google Places, Routes, and Weather APIs
- Built a real-time rerouting engine for venue, time, and duration edits on one shared core, with a locked-time invariant that never alters confirmed past stops
- Hardened it with 145 automated tests (16 Playwright end-to-end, 129 unit), including a stubbed lying model that proves code always overrides bad model output; wrote a 10-page business plan toward Y Combinator
Garbage Sorting Model | Python, TensorFlow, Keras, MobileNetV2, YOLOv11, OpenCV
2026- Built a real-time computer-vision system that sorts items on a live webcam feed into compost, garbage, recycling, or none, using MobileNetV2 transfer learning
- Curated about 12,000 images from 11 public datasets with an isolated 70/15/15 train, validation, and test split, reaching 95.9% validation accuracy
- Diagnosed a 2.5× recycling class imbalance that skewed live predictions; rebalanced the data and added YOLOv11-nano detection with a 50% confidence gate to isolate the object from background clutter
- Ran inference at about 20 fps in an OpenCV loop, with a tracking box and per-class confidence shown on screen
STARR Patch | Arduino, C/C++, Physiological Sensors
2025 to Present- Developed a prototype SpO2 and blood-pressure sensing system, integrating sensors with a microcontroller for real-time acquisition
- Wrote firmware to initialize, sample, filter, and validate noisy raw sensor readings
- Built serial and live-plotting tools to isolate faults across sensor configuration, wiring, timing, and signal processing
BreadBot | Fusion 360, Rotary Encoders, Microcontrollers
- Co-designed a topology-aware bread slicer that measures a loaf’s surface profile with a rotary-encoder tracking mechanism and adapts the cutting path to it
- Modelled the structure, track, and mounts in Fusion 360, iterating through physical tests to reduce the mechanical play, friction, and misalignment that skewed encoder readings
- Chose contact sensing with an encoder over computer vision, keeping hardware, calibration, and compute simple
GaragePoker | React, TypeScript, CSS
2026- Built a browser poker table (garagepoker-flax.vercel.app) that looks like a modern online poker client, using only HTML and CSS rather than a 3D engine
- Separated game-state logic from presentation, so a single source of truth drives seats, cards, chip stacks, bets, the pot, and action controls
Personal Website | HTML, CSS, JavaScript
2026- Designed and built this site from scratch with no framework: an animated business card that unfolds into an interactive painted room
- Drew every object in the room in CSS and inline SVG, fitted to the painting’s perspective with computed homography transforms, with a live GitHub contribution grid and content-driven panels
Technical Skills
- Languages
- Python · TypeScript · JavaScript · C/C++ · HTML/CSS
- Frameworks & Libraries
- React · Next.js · TensorFlow · Keras · OpenCV · YOLO · NumPy
- Developer Tools
- Git & GitHub · Vercel · Playwright · Upstash Redis · Fusion 360 · AI-assisted development
- Hardware
- Arduino · microcontrollers · sensor integration · rotary encoders · serial communication · prototyping