I turn raw, messy data into systems that work - ML pipelines, analytical dashboards, and decision-ready outputs that hold up under scrutiny.
I'm a Data Science student at Simon Fraser University with a deep curiosity for how numbers tell stories. My work sits at the crossroads of statistical thinking and creative problem-solving — finding patterns where others see noise.
Before diving into data full-time, I spent 3+ years at Amazon where I learned to think in systems — monitoring KPIs, spotting bottlenecks, and optimizing workflows at scale. That operational mindset now shapes how I approach every dataset.
My coursework spans Data Structures, Probability & Statistics, and Software Engineering. When I'm not wrangling data, you'll find me building side projects, competing on Kaggle, or sketching ideas before they become code.
Coursework: Data Structures & Algorithms, Probability and Statistics, Introduction to Software Engineering, Introduction to Computer Systems, Artificial Intelligence, Linear Algebra, Discrete Mathematics
A machine learning pipeline that classifies financial transactions not just as fraud or legitimate, but by how urgently they need to be investigated. Built for HackML 2026 on 6.2 million real-world transactions with extreme class imbalance.
22 years of Vancouver Open Data across 24 neighbourhoods, queried in SQLite and visualised with ggplot2 faceted charts. The Central Business District accounts for 30.3% of all incidents — but counts without population denominators are not the same as risk.
An end-to-end analytics pipeline for 16 years of BC Ministry of Health surgical data — 202,953 rows cleaned with 10 documented decisions, 5 audience-specific SQL views in DuckDB, and a Power BI dashboard that surfaces the P90 story the median always hides.
A physics-grounded scoring engine that takes five climate inputs and recommends Solar, Wind, or Hydro — with full point breakdowns and a confidence margin. Validated against real NOAA station data using ANOVA and permutation feature importance across a 3000+ line R pipeline.
A solver that finds near-optimal solutions for any Rubik's Cube scramble in under 100ms, with a 3D animated interface that plays back every move step by step. Built using Kociemba's Two-Phase Algorithm with precomputed pruning tables.
Python
SQL
R
Pandas
NumPy
dplyr
Matplotlib
Seaborn
ggplot2
Excel
Sklearn
XGBoost
SMOTE
K-Means
DBSCAN
Rnd Forest
Git
Docker
FastAPI
React
Java
Streamlit