Overview
A financial decision-support system that automates SME loan underwriting. It acts as an AI Financial Analyst that visualizes corporate risk architecture.
Tech Stack
Python (Streamlit)Google Gemini 2.0 FlashSQLAlchemy & SQLiteGraphvizDuckDuckGo Search API
Features
- Fraud Cross-Validation: Automatic flagging of discrepancies between Bank Statements and Audit Reports.
- Risk Architecture Mapping: Visual red/green node cues indicating financial health.
- Market Pulse: Real-time sentiment analysis via web search integration.
- Credit Memo Generator: Auto-generation of formal approval PDFs.
Key Decisions
01
Implemented a strictly grounded RAG system that forces the AI to provide page-number citations from the uploaded audit PDFs.
02
Chose Graphviz over standard D3.js to enable fast, server-side generation of complex corporate hierarchy maps directly within Streamlit.
Learnings
→Developed a robust "Variance Detection" algorithm that mathematically flags inconsistencies between reported revenue and actual bank inflows.
→Learned to automate the creation of professional PDF credit memos using the fpdf library based on structured AI verdicts.