GenAI Risk & Compliance Analyst for Financial Filings

Overview

This project is a GenAI-powered assistant that analyzes SEC 10-K filings and other financial documents to extract key risks, litigation, and compliance issues. It uses Retrieval-Augmented Generation (RAG), tracks LLM prompts/responses with MLflow, and provides an interactive Streamlit UI for analysts.

Key Features

• Ingests real SEC filings and sample filings
• RAG with FAISS for document retrieval
• LLM (Hugging Face) for risk extraction and summarization
• MLflow for prompt/response tracking
• Streamlit interface for interactive Q&A
• Prompt templates for risk/compliance

Architecture

graph TD
    A[SEC Filings/Docs] -->|"Ingestion"| B[Chunking & Embedding]
    B -->|"Vectors"| C[FAISS Vector Store]
    C -->|"Retrieval"| D[Relevant Chunks]
    D -->|"Prompt"| E["LLM (Hugging Face)"]
    E -->|"Answer"| F[Streamlit UI]
    E -->|"Tracking"| G[MLflow]

Loading

Project Structure

• ingestion/ - Data ingestion scripts
• rag/ - RAG and vector store logic
• llm/ - LLM and prompt templates
• mlflow_tracking/ - MLflow tracking server, logs, and setup scripts
• airflow/ - Airflow DAGs for pipeline orchestration
• api/app.py - Streamlit app (main entry point)
• data/sec/ - Downloaded and sample SEC filings

Getting Started

Option 1: Run End-to-End Pipeline with Airflow

1. Install dependencies: pip install -r requirements.txt
2. Initialize Airflow:

   export AIRFLOW_HOME=$(pwd)/airflow
   airflow db init
   airflow webserver -p 8080 &
   airflow scheduler &

3. In the Airflow UI (http://localhost:8080), enable and trigger the genai_risk_pipeline DAG for end-to-end orchestration (ingestion, embedding, LLM, MLflow, Streamlit).

Option 2: Manual Steps (No Airflow)

1. Install dependencies: pip install -r requirements.txt
2. Start MLflow tracking server:

   bash mlflow_tracking/mlflow_setup.sh
   # or manually:
   mlflow server --backend-store-uri sqlite:///mlflow_tracking/mlflow.db --default-artifact-root ./mlflow_tracking --host 0.0.0.0 --port 5000

3. Ingest SEC filings: python ingestion/sec_ingest.py
4. Embed and store: python rag/embed_and_store.py
5. Run the Streamlit UI:

   streamlit run api/app.py

6. Use the UI to search for litigation, risk, and compliance issues in filings

Option 3: Docker (Recommended for Production)

A. Single Container (Streamlit + MLflow)

1. Build the image:

   docker build -t genai-risk-analyst .

2. Run the container:

   docker run -p 8501:8501 -p 5000:5000 genai-risk-analyst

3. Access Streamlit at http://localhost:8501 and MLflow at http://localhost:5000

B. Full Orchestration with Docker Compose (Streamlit + MLflow + Airflow)

1. Start all services:

   docker-compose up --build

2. Access the UIs:
   • Streamlit: http://localhost:8501
   • MLflow: http://localhost:5000
   • Airflow: http://localhost:8080
3. In Airflow, enable and trigger the genai_risk_pipeline DAG for end-to-end orchestration.

Example Use Cases

• List potential litigation risks in a 10-K
• Compare debt covenants across quarters
• Summarize ESG compliance sections
• Detect unusual spikes in transaction logs

---