CashFlow AI
- Role —
- Independent builder
- Stack —
- Python, FastAPI, Streamlit, SQLite, scikit-learn
CashFlow AI is a modular, local-first project for turning transaction data into useful financial signals without treating a raw statement as a black box.
The project is structured around explicit ingestion, categorisation, recurrence detection, forecasting and anomaly-detection stages. Its design separates transaction category from financial role so that transfers, refunds and withdrawals are not misreported as income or expense.
Focus areas
- Typed data boundaries and auditable source handling
- Chronological evaluation for forecasting rather than shuffled time-series splits
- Clear treatment of uncertainty and gaps in financial data