Financial Statement Analysis Pipeline

Overview

This project automates the extraction, processing, and analysis of financial statements using GenAI. It processes financial PDFs, extracts key financial data, and generates structured CSVs and a final summary report.

The process, justification, and future recommendations the final pipeline is documented in Project Documentation.md (different to README), and the raw jupyter notebook is in exploring.ipynb.

Smashed their interviews, still got rejected because the role they wanted to make for me, they didn't have headcount or something? Not sure. Past life. Anyways.

Pipeline Workflow

graph TD
    A[PDF] --> B[Extract]
    B --> C[Preprocess]
    C --> D[List of Tables]
    D -->|Send to GenAI| E[Extract Key Financial Metrics]
    
    E -->|Save| F[GenAI CSV Tables]
    F --> H[Final CSV Output]
    F -->|Analyze| G[Generate Summary Report]
    
    G --> I[Final Summary Report]

Loading

Project Structure

📁 financial_extractor/
│── 📁 data/               # Extracted CSV files are placed here
│── 📁 pdf_inputs/         # PDFs for processing
│── 📁 reports/            # Final summary reports
│── 📁 scripts/            # Contains all processing scripts
│   │── preprocess_data.py    # Extracts and cleans text from PDFs
│   │── generate_tables.py    # Processes tables via GenAI & saves CSVs
│   │── genai_summary.py      # Generates financial summaries
│   │── config.py             # Loads API keys, etc.
│── .env                  # API keys and config variables
│── main.py               # Runs the full pipeline
│── requirements.txt       # Python dependencies
│── Project_Documentation.md       # Project documentation & reasoning
│── README.md            

Setup Instructions

1. Install Dependencies

First, install the required dependencies:

pip install -r requirements.txt

2. Add Your API Key

Create a .env file in the root directory and add your Gemini API key:

GEMINI_KEY=your_api_key_here

3. Run the Pipeline

Execute the main script to process financial PDFs and generate reports:

python main.py

How It Works

1. Extract Financial Data from PDFs

• Uses pdfplumber to extract raw text from structured financial reports.
• Cleans and preprocesses the text to remove irrelevant headers/footers.
• Splits the text into different financial sections (e.g., Income Statement, Balance Sheet).

2. Process Tables Using GenAI

• Each segmented financial section is sent to Gemini AI for structured extraction.
• AI returns structured JSON financial data.
• Data is formatted into CSV tables and stored in /data.

3. Generate a Summary Report

• Reads all extracted CSVs and processes key financial trends.
• Uses GenAI to generate a Markdown financial summary.
• Converts the summary to a PDF report.

Outputs

✅ Extracted Financial Tables → /data/ ✅ Summary Report (PDF) → /reports/summary_report.pdf