Built for the IIT Hyderabad National AI/ML Hackathon 2026 - Top 10 Finish out of 7,700+ participants

IntelliCredit is an end-to-end credit analysis pipeline that reads raw financial documents (Annual Reports, Bank Statements, GST Filings) uploaded by Indian MSMEs and NBFCs, extracts key financial metrics, and generates a structured credit health dashboard.

• Multi-format document ingestion: PDF (bordered + borderless tables), CSV, Excel, DOCX
• OCR fallback for scanned documents with image preprocessing
• AI-powered metric extraction via Claude API (Anthropic)
• Offline fallback engines: Regex Analyzer + Derived Metrics Engine
• Supports Indian financial format: lakh/crore unit normalization, parentheses negatives
• Computes: Revenue, PAT, EBITDA, Net Worth, Total Debt, D/E Ratio, DSCR
• Estimation mode from raw Bank Statement and GSTR-3B data when no annual report is available

intellicredit/
├── app.py                  # Main application entry point
├── core/
│   ├── data_ingestor.py    # Multi-format document parser (PDF/CSV/Excel/DOCX)
│   └── ai_analyzer.py      # Financial metrics extractor + derived metrics engine
├── static/                 # Frontend assets (if applicable)
├── templates/              # HTML templates (if applicable)
├── docs/                   # Reference documents and additional notes
├── requirements.txt        # Python dependencies
├── .env.example            # Environment variable template
└── .gitignore

• Python 3.10+
• pdfplumber, PyMuPDF (fitz), pytesseract — PDF parsing
• pandas - structured data handling
• Anthropic Claude API - AI-powered extraction
• XGBoost + SHAP - credit scoring and explainability
• Pillow - OCR image preprocessing

git clone https://github.com/HarshApurva/intellicredit.git
cd intellicredit

python -m venv venv
source venv/bin/activate   # Windows: venv\Scripts\activate

pip install -r requirements.txt

cp .env.example .env
# Add your Anthropic API key inside .env

python app.py

1. User uploads financial documents (Annual Report PDF, Bank Statement CSV, GSTR-3B)
2. DataIngestor routes each file to the correct parser based on file type
3. Extracted text and tables are passed to FinancialMetricsExtractor
4. If Claude API is available: Claude agent extracts metrics via structured JSON output
5. If Claude API is unavailable: Offline regex engine + derived metrics estimation kicks in
6. Dashboard displays: Revenue, PAT, EBITDA, Net Worth, Total Debt, D/E Ratio, DSCR

Built by Team SPB for the IIT Hyderabad National AI/ML Hackathon, March 2026.

MIT License — Copyright (c) 2026 Harsh Apurva, Aditya Sharma, Aditya Bakhale

Permission is hereby granted, free of charge, to any person obtaining a copy of this software to use, copy, modify, merge, publish, distribute, and/or sublicense it, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the software.