machine learning
computer vision is the part of ml that keeps pulling me in: scenes, objects, faces, products, and the meaning hidden inside a frame.
during my co-op at the Ontario Public Service, i built AWS Bedrock RAG and LLM workflows that reduced analyst triage time by 35%.
that work made my direction clearer: research and development of vision models for scene understanding, affordances, and real-world context.
i finished Fanshawe's Artificial Intelligence co-op program with a 4.2 GPA, focused on deep learning, generative ai, deployment, and mlops.
- i want to work on perception systems people can trust under messy lighting, partial views, and real-world uncertainty.
- my stack includes Python, PyTorch, OpenCV, FAISS, LangChain, LangGraph, AWS, FastAPI, Docker, and backend systems.
things i built
my projects usually start with one question: what can a machine understand from imperfect input?
- ona is my Object Navigation Assistant, turning camera input into object, depth, face, OCR, and scene state. demo
- friendnet is a custom vision model for friendly-face classification, compared against ResNet and CLIP baselines. demo
- copnow taught me product gravity. i co-founded the marketplace and led ML/backend work around trust, commerce, analytics, and APIs. demo
- orchestra is a Rust/macOS synchronization daemon for AI coding workspaces: deterministic templates, atomic writes, and drift detection.
before that
before ml became the center, i was mostly a backend engineer. at SHIIP, i built high-throughput Go and Python services for logistics.
earlier, i worked on practical web systems at Curacel and TEEK-TECH.
my computer science thesis was a sensor-guided autonomous vehicle. that project still explains me: software meeting the physical world.