I build systems that turn messy data into useful decisions, bridging business intelligence and production ML deployment.
I started in geology, where you learn to reason carefully from incomplete evidence. Now I do that every day with data: executive dashboards in Excel and Power BI, validation-first fraud and risk investigations, churn models that ship behind a REST API, and the static portfolio site that presents all of it.
- Portfolio: elishabassey.netlify.app
- Contact: basseyelisha99@gmail.com
| Project | What it is |
|---|---|
| customer-churn-retention | Logistic regression churn model (test ROC-AUC 0.858) served through a FastAPI endpoint, plus the Excel executive dashboard for the same 7,043-customer dataset |
| DataDNA-Challenge | Monthly #DataDNA challenge builds: a validation-first PayPesa wallet-fraud audit (every brief hypothesis formally tested) and Golden Wok delivery intelligence in Power BI |
| zion-finance-data-challenge | Reproducible Python pipeline and 15-sheet Excel workbook tracing $2.84M of failed fee revenue in East African payment data |
| sunrise-coffee-performance | Two-year performance review of 60 UK coffee shops: regional margin, loyalty value, revenue mix and seasonality, rebuilt from five raw CSVs |
| airbnb-market-intelligence | Airbnb market analysis with documented statistics and a full Power BI build behind it |
| CRM-Sales-Analytics | ML-powered REST API for sales pipeline prediction |
| TrustLake | Deterministic data validation with AI-assisted explanations |
| My_Portfolio | The static site behind my portfolio: case-study pages with business and technical documentation, generated by a small Python build |
- Verify every number against the raw data before it gets published, then document the trail
- Report effect sizes, not just p-values; label metrics for what they actually measure
- Write documentation that reads like a person, not an auditor
- Correct my own numbers in public when a re-check proves them wrong
Every case study above has a business page and a technical page on the portfolio, where the reasoning is laid out end to end.