JohnApollos/FinLens-AI

Intelligent financial document analysis engine - anomaly detection, ratio analysis, and AI-generated risk narratives for African and emerging market financial statements. # FinLens AI — Intelligent Financial Document Analysis and Risk Intelligence Engine --- ## Section 1 — The Problem Most financial risk analysis still happens in spreadsheets and email threads. An analyst reads a PDF, manually extracts numbers into Excel, computes ratios, compares against remembered benchmarks, and writes a narrative from scratch. For a single document this takes hours. For a portfolio of documents it becomes a bottleneck. A critical challenge in emerging markets is that most financial analysis AI tools are trained exclusively on Western datasets (S&P 500 / SEC). When applied to African financial statements (like Kenyan listed companies), these models systematically misclassify healthy entities as anomalies because they operate under different capital structures, liquidity dynamics, and regulatory guidelines (e.g., CBK banking regulations versus US Federal Reserve rules). A tool trained on S&P 500 companies will misread a Kenyan bank's balance sheet, misinterpreting high liquid asset buffers or regional non-performing loan structures as default warnings. --- ## Section 2 — What FinLens AI Does FinLens AI automates the entire risk review process: upload a financial document, receive a structured risk assessment with anomaly flags, industry benchmark comparisons, and a plain-English risk narrative in under 60 seconds. Key features include: - **Intelligent Ingestion**: Hybrid text extractor using Google Gemini 1.5 Flash (JSON Mode) with deterministic double-entry validation equations ($Assets = Liabilities + Equity$) and local regex-based heuristics fallbacks. - **High-Performance MD5 Caching**: Programmatically hashes uploaded files via MD5. Duplicate renders, page tab switches, or PDF compilation requests load instantly from the JSON cache in **under 0.01 seconds**, saving API quotas. - **Solvency Math**: Auto-computes Altman Z'-Scores (manufacturing) and Z''-Scores (service/banking/fintech) with color-coded distress rating zones. - **Explaina …