Full-Time and Internship Experience
💳 Stripe - ML Exploration
Leading Stripe’s ML library; components for preprocessing, pre-built models, experimentation, evaluation, deployment. Re-engineered models to estimate network costs in native XGBoost/Flyte, reducing train time by 10x, resource use by 50%. Coordinated migration of 44+ ML workflows from 14 teams, directly onboarding 6 teams onto Flyte's workflow engine. Measured and evaluated task/pod spin-up latency for ML workflows by introducing new SLOs on Prometheus/Grafana.
💸 Wealthsimple - ML Products and Platform
Built productivity/analytics AI assistant with code and retrieval augmented generation, LangChain agents, Mistral/Llama. Open-sourced privacy-preserving web portal/API to safely interact with LLMs (TypScript/FastAPI/ChatGPT/Cohere). Won company hackathon by delivering lead scoring web app to VP Growth using LGBM regression, Streamlit. Drove secure web platform/services, accelerating the creation of data apps from months to minutes (Okta/OAuth/Docker). Headed a Python package to preprocess, train, and deploy model variants to MLflow/Triton for evaluation and A/B testing. Doubled Python/Rasa chatbot helpfulness for ~2K daily chats by training entity/intent classification, streamlining NLP rules. Designed 3 tools to semantically search similar clients, portfolios, chats, with embeddings, k-nearest neighbors, Elasticsearch. Saved ~$510K in people/software costs by leading tooling to automigrate company BI platform from Periscope to Superset. Developed Airflow pipeline to validate all $50B assets under management across Oracle, AWS Redshift data stores.
👤 Meta/Facebook - Ads Ranking ML
Optimized ad auction bids shown to 3B+ users by redesigning deep learning models to better predict return on ad spend. Added multi-task learning modules to FB's core ML library to suggest bids given conversion probabilities (PyTorch/Caffe). Mitigated 30% data loss in Hive pipelines by imputing missing data with regression, eliminating outliers, feature engineering. Introduced experimentation API to evaluate model performance on metrics such as Kendall's Tau (Python/Pandas). Identified model drift between online/offline performance, researched and proposed solution, drove alignment with 2 teams.
🛒 Shopify - Inventory Logistics
Powered core inventory insights for 1.75MM merchants with data modeling and analytics workflows (Python/SQL/Spark). Fixed failing Spark ETL pipeline and reduced compute time by ~60% to efficiently surface key metrics across shop locations. Spearheaded new dashboard, Kafka events, and wrangled 7+ Presto tables to assess engagement with cross-platform features. Advised strategic product decisions by analyzing HTTP requests between Shopify APIs and third-party apps in Splunk logs.
🌐 Orbis Investments - Global Web Portal
Launched service-oriented client-facing web apps to view $39B in funds with TypeScript React, ASP.NET Core, SQL Server. Saved developers 150+ manual clicks per daily deployment by creating a Swagger endpoint to instantly populate the DB. Implemented data tables and dynamic widgets with contracts, RabbitMQ message handlers, API endpoints, front-end code. Architected an API microservice and altered the corresponding web admin dashboard to manage user permissions.
💵 Canalyst Financial Modeling - Equity Research Tooling
Engineered C#/.NET tools to clean and screen data across 4K+ quantitative models used by 500+ investment professionals. Improved data quality by 53% through an algorithm to extract and structure textual data from financial reports. Pioneered a Plotly/Flask dashboard to help users visualize, analyze, and compare company KPIs across sectors. Wrote software to manipulate tabular data and Excel formulas using parse trees, regex, LINQ, test-driven development. Eliminated toil by automating Google Sheets logging and refactoring classes to surface detailed error/solution messages. Diagnosed and shipped frequent bug fixes and rapid support for 80+ equity analysts during peak usage periods.
👨💻 Tech Bytes Foundation - Data/AI/ML Education
Led the design and development of 10 technical projects for an 8-week curriculum in data science, AI, and ML, using technologies such as Pandas, NumPy, Scikit-learn, OpenAI Gym, Keras. Managed tasks with Agile/Jira/Confluence, trained new developers and presented at technical workshops.
Other Experience
👨🏫 UBC Department of Computer Science - Teaching Assistant
Led weekly labs on systematic program design (CPSC 103/107) for ~60 students, coordinated with professors Meghan Allen and Steve Wolfman, and evaluated student code. Concepts taught include: data analysis, graphs, trees, recursion, search, dispatch, and functional programming. Technologies taught include: Python, Racket, Jupyter, Conda, NumPy and Matplotlib.
🌱 Launch Pad Student Design Team - Android
Built Java/Kotlin classes in Android Studio for Soil TopARgraphy: a teaching aid that analyzes soil distribution through AR. Collaborated with 5 developers over GitHub and ZenHub. App presented at Multidisciplinary Undergraduate Research Conference and International Soils Meeting (San Diego).
🗣 Fostering Debate Talent - Speech and Debate Coach
Coached students of all ages to win awards on the international speech and debate circuit, including at tournaments such as Harvard and Stanford.
Education
🎓 Master's, Computer Science with ML Specialization - Georgia Tech (4.0/4.0)
In progress.
🎓 Bachelor's, Honors Computer Science and Business - UBC (4.30/4.33)
Top GPA in Program, Trek Excellence Scholar (awarded to the top 5% of students in their program), Dean's Honor List, BC Achievement Scholarship.
📜 Advanced Courses
Natural Language Processing, ML for Trading, AI for Robotics, AI Ethics, AI I/II, Applied ML, Algorithms, Software Engineering, Databases, Information Systems, Financial Modeling, Operations and Logistics, Economics.