Skip to content
All projects

Statement IntelliScan

Extracts transactions from password-protected bank statement PDFs into CSV and predicts transactional risk profiles with ML models tuned for highly imbalanced data.

Role
Developer
Organization
Personal project
Period
Jul 2025 – Oct 2025
Type
Web app & machine learning

Highlights

  • FastAPI extraction endpoints with a Next.js backend layer
  • pdfplumber parsing and cleaning of password-protected statements into tabular CSV
  • TF-IDF, transaction-volume and temporal-window features for Logistic Regression and MLP models
  • GroupKFold cross-validation to prevent cross-file leakage; PR-AUC with threshold tuning