An AI-powered solution to automate Anti-Money Laundering (AML) workflows using modular, specialized agents built on CrewAI.
Winner of the 1st Place Prize at the Agentic AI Hackathon conducted by Techvantage.ai in collaboration with CrewAI.
Manual AML compliance processes often result in delays, human errors, and overlooked red flags. This project offers a scalable and explainable multi-agent AI system to monitor, detect, and manage suspicious financial transactions efficiently.
Using CrewAI for orchestration, each agent is tasked with a specific AML responsibility — from prediction to validation and reporting — enabling a seamless and autonomous compliance pipeline.
- Agentic Workflow Orchestration: CrewAI coordinates task-specific agents for streamlined AML operations.
- Fraud Detection: Uses XGBoost-based models to identify potentially fraudulent transactions.
- Compliance Validation: Validates entities against global sanctions and PEP (Politically Exposed Persons) databases.
- Explainable Reporting: GPT-4o and SHAP explain model decisions with rich markdown summaries.
- Case Management: Assigns risk-based priorities and tracks flagged cases.
- Frontend Interface: A user-friendly Streamlit dashboard for seamless interaction.
- AI & Agents: CrewAI, GPT-4o, Pydantic
- Machine Learning: XGBoost, SHAP, Scikit-learn
- Visualization: Streamlit, Plotly, Seaborn, Matplotlib
- Data Handling: Pandas, NumPy, Joblib, JSON, Markdown
- Environment: Python 3.11, dotenv
-
Clone the Repository
git https://github.com/VenkataSakethDakuri/AML_Crew
-
Set Up a Virtual Environment
python3.11 -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate
-
Install Dependencies
pip install -r requirements.txt
-
Run the Streamlit App
streamlit run app.py
Watch the full walkthrough on Loom: Demo Video
- Precision (Fraudulent): 0.82
- Recall (Fraudulent): 0.93
- F1-Score (Fraudulent): 0.87
Model effectively detects laundering with high recall while reducing false positives.
- Graph-based anomaly detection
- Real-time sanctions/PEP data updates
- Custom risk modeling for institutional compliance
MIT License. See LICENSE for details.
Crafted for the Techvantage.ai BFSI Hackathon using CrewAI. ✨