Payment Recovery ML (End-to-End System)
FlagshipPredicts which failed / unpaid transactions are likely to be recovered and prioritises outreach by expected recovered revenue.
- End-to-end ML pipeline (SQL → features → model → Streamlit app)
- Calibrated Logistic Regression with PR AUC, Brier Score, lift analysis
- Expected value ranking: amount × probability to maximise recovered revenue
MLFinanceStreamlitSQL
Finance Collections & DSO Forecasting
Time series forecasting for collections and Days Sales Outstanding (DSO), helping finance teams anticipate cash flow and risk.
- Built using Prophet and invoice / AR data
- Produces visual forecasts of expected cash inflows and DSO trends
- Supports FP&A and treasury planning with scenario-style views
Time SeriesFinanceProphet
SQL Analytics & BI Dashboards
SQL-based analytical views and dashboards for AR, collections, and operations, consumed in Tableau / Power BI.
- Heavy use of window functions, cohort logic, and aggregations
- Aging buckets, collection performance, and payment behaviour segmentation
- Designed for self-service exploration by non-technical stakeholders
SQLBIFinance
More projects (e.g. SQL notebooks, BI dashboards, and smaller experiments) are available on GitHub.
