About Me
Data Science student at the University of Waterloo with an interest for machine learning, web development, and solving complex problems.
Honours Data Science, BMath
University of Waterloo
Sep. 2024 - May 2029 (Expected)
Relevant Coursework
Activities
- Data Science Club
- Applied Math, C&O Club
- Various Hackathons + Datathons
Experience
AI / Software Engineer
IncomingPractice Better
- >Built semantic search systems with Azure OpenAI text-embedding-3-large and MongoDB, enabling support agents to surface similar resolved tickets across a 3,000+ ticket knowledge base and letting practitioners unlock insights from thousands of client session histories via hybrid semantic search
- >Developed the custom charting pipeline for an LLM-powered auto-charting feature that generates structured patient charts from session notes, using a chain handler pattern to orchestrate multi-step agent workflows and support practitioner-defined chart templates
- >Developed an LLM-powered ICD-10 diagnosis code classifier spanning 70,000+ codes that suggests codes from session note summaries, achieving 83% F1 score and reducing manual lookup time for practitioners
Data Analyst, Supply Chain
Celestica
- >Trained a generative AI model with OCR to extract key product specifications — including case size, resistance, and maximum voltage — from technical drawings, reducing manual data entry
- >Leveraged APIFY to scrape part data into JSON, then performed ETL with Python and Pandas. Increased the proportion of benchmarked parts by 18% and total dollar value by 22%
Machine Learning Research Assistant
Fluid Flow Physics Group
- >Developed a Random Forest regression model to predict fundamental frequency and sound pressure level, optimizing hyperparameters with cross-validation and generating contour plots to visualize model performance
- >Authored Python scripts to transform vector data from VTU-formatted files into Pandas data frames then export as CSV for further analysis
Memory Archive
Recovered data fragments from previous constructs. Each memory represents a unique solution to complex problems.
Argus — Arctic Defence Platform
Full-stack geospatial dark-vessel detection platform that correlates SAR satellite imagery with live AIS tracking to surface suspicious ships in Arctic waters
Red Light Detection — Raspberry Pi CV
On-device YOLOv5-nano traffic-light classifier running at 12+ FPS on a Raspberry Pi 4 with Pi Camera, hitting 94% mAP on a custom 3,000-image dataset
SleepyPlays
Automated YouTube Shorts content creation system using AI for story generation, voice synthesis, and scheduled video publishing
MealMinder
Smart meal planning and grocery management application with AI-powered meal plan generation