Skip to content
View ayokumo's full-sized avatar

Block or report ayokumo

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ayokumo/README.md

Ayokunmi Lawal

Applied Mathematics student at York University, Toronto. I build the whole path from raw data to decision: ingestion, a tested warehouse, a model, and a dashboard someone actually uses.

Currently: Sports Analyst Intern, York University Athletics. Varsity football game databases in Python and SQL, player performance dashboards in Power BI for coaching staff.

Open to: Winter 2027 (Jan to Apr) or 8 month (Jan to Aug 2027) co-op in data, analytics, or actuarial roles.


Featured work

Question: Do Bank of Canada rate changes show up in Toronto shelter demand, and can it be forecast? Built: ELT over five public sources into DuckDB, dbt star schema on a 433,362 row fact table, 26 automated tests. Found: Occupancy rose about 0.64pp per 1% rate increase at a 6 month lag. SARIMA and Prophet both lost to a naive baseline over 37 fold rolling origin CV, so the forecast layer was cut rather than shipped. Python SQL dbt DuckDB statsmodels

Question: What actually predicts overdose incidence across Toronto's 158 neighbourhoods? Built: Geocoded 2018 to 2025 incidents, spatial diagnostics, spatial lag and error models, then Poisson, Random Forest, and Gradient Boosting under repeated k fold CV. Found: Low income prevalence drives most of the clustering. Model error fell from 52 to 33 incidents, and the remainder traces to individual addresses below neighbourhood scale. Python geopandas PySAL scikit-learn

Question: Can a transparent model produce daily predictions at zero cost? Built: Scheduled GitHub Actions pipeline (daily, hourly on match weekends, 15 minute checks near kickoff) feeding a weighted factor model and a live dashboard on GitHub Pages. Ingestion fails safe by keeping prior data. Python GitHub Actions REST APIs


Stack

Analysis: Python (pandas, NumPy, scikit-learn, statsmodels, geopandas), SQL, R Modelling: OLS and Poisson regression, Random Forest, Gradient Boosting, SARIMA, Prophet, spatial regression, cross validation Data: dbt Core, DuckDB, PostgreSQL, ETL and ELT design, automated data quality tests Reporting: Power BI (DAX, Power Query), Excel, Tableau, Streamlit Tooling: Git, GitHub Actions, Docker, pytest


Contact

LinkedIn · Portfolio · lawalayokunmi25@gmail.com

Pinned Loading

  1. toronto-overdose-spatial toronto-overdose-spatial Public

    Geospatial analysis and spatial regression of suspected non-fatal opioid overdoses in Toronto shelters, 2018–2025

    Jupyter Notebook 1

  2. ayokumo ayokumo Public

  3. toronto-housing-project toronto-housing-project Public

    End-to-end data platform linking macro-financial indicators (BoC policy rate, rental market) to Toronto shelter demand with dbt star schema, OLS regression, SARIMA/Prophet forecasting with rolling-…

    Python

  4. pl-predictor pl-predictor Public

    Python