BACK TO CAREER

03 / PORTFOLIO

Selected projects

Personal projects in statistical simulation, data workflows, and practical productivity systems.

PROJECT_01Data pipeline and interactive visualization
Toronto Rent Heatmap Lab interface with filters, regional map, and rent metricsLIVE MAP

Toronto Rent Heatmap Lab

A city data workflow and interactive map prototype for comparing rent pressure across Toronto.

WHAT I BUILT

  • Built a Python workflow for data extraction, transformation, and loading from normalized and CMHC style tables, including handling for suppressed rent values.
  • Designed a SQL dimensional model and Power BI ready exports for rent, vacancy, unit count, geography, and year.
  • Built a GitHub Pages choropleth with official City of Toronto Community Council boundaries. The current map uses sample rent values.
PROJECT_02Statistical simulation and risk analysis
Stock Monte Carlo Forecaster showing scenario paths and downside risk metricsLIVE APP

Stock Monte Carlo Forecaster

A Python tool for exploring stock price uncertainty through reproducible GBM simulations and downside risk metrics.

WHAT I BUILT

  • Estimated drift and volatility from historical prices, then generated seeded GBM paths for repeatable scenario analysis.
  • Calculated percentile ranges, VaR, CVaR, probability of loss, and target hit probability.
  • Packaged the work as reusable Python modules with a command line interface, Plotly charts, and a Streamlit app.
PROJECT_03Workflow and productivity prototype
Career OS repository documentation showing the working local prototype scopePROJECT RECORD

Career OS

A local Electron prototype for organizing a personal co-op search from job review to application package tracking.

WHAT I BUILT

  • Modelled collected jobs, review decisions, package folders, application stages, and preference rules in one local workflow.
  • Used Electron, JavaScript, and JSON backed state with a limited preload bridge for local file operations.
  • Defined the product need, workflow, priorities, and review decisions, then used an AI coding agent for a substantial share of implementation and iteration.