Sentinel AI – AI-Powered Mule Account Detection

Detect suspicious mule accounts faster with AI-driven risk assessment, automated investigation workflows, and human validation.

Comment

Inspiration

My sister had been working in Bank of Baroda as Assistant Branch Manager. One time she had a mule account case where the money was flowing in and out of that same account at a very big volume, around 12:30AM. It was the first time I had heard the word “Money Mule.” I don't want to say who the branch is for privacy reasons. I was surprised that they didn't have a solution that specifically detected and investigated these accounts proactively. This occurred during AI's transformative era in 2025. So I thought, What's the point of investigators continuing to do so much manual work when AI and intelligent workflow automation can help? The question sent us down a path to develop a mule investigation platform called Sentinel AI that could be powered by Artificial Intelligence and human beings, rather than by humans alone.

What it does

Sentinel AI helps investigators detect and investigate suspicious mule accounts involved in financial crimes such as money laundering. The platform's AI risk assessment service performs a risk analysis of an account and provides fraud indicators, including a score, flags for suspicious activity, and confidence scores. UiPath Maestro manages all the investigation process, automating the AI analysis and providing the results for manual review. Investigators can subsequently either approve or reject the AI recommendation, ensuring that crucial financial decisions are always made by a human.

How we built it

We have four integrated components in our solution:

React + TypeScript for investigator dashboard & case management UI. Express.js backend with REST API's and investigation cases. Providing AI-powered mule account risk scoring using FastAPI (Python).

  • UiPath Maestro orchestrating the investigation workflow, calling external AI services, processing results and managing the Human Approval process.

All components communicate with each other by means of REST APIs, enabling each component to operate in isolation and create a complete investigation platform.

We encountered a number of issues.There were some problems we faced.

One of our greatest challenges was bringing all the technologies together, so that they would form a cohesive whole.

We had to:

Link React front end to Express back end. Include an external FastAPI AI service. Configure UiPath Maestro to make API calls, deserialize Json response, and initiate and orchestrate workflow executions. Develop a Human Approval interface that is clear, simple and intuitive to present AI findings.

The following are things that we are proud of that we have achieved.Here are accomplishments we are proud of:

Successfully developed end-to-end platform which investigated mule accounts with AI assistance. Incorporated UiPath Maestro with external AI services through REST APIs. Designed a professional investigator dashboard to manage cases. Adopted a human-in-the-loop approval process to promote responsible use of AI. Illustrated how AI, workflow orchestration and human decision-making can complement each other to speed financial crime investigations.

What we learned

This project was a great learning curve on much more than just creating single applications. How workflow orchestration, AI services, back end APIs and modern front end applications can be combined in an actual solution. We were able to gain hands on experience of using UiPath Maestro; FastAPI; and REST API integration as well as human centred AI design. The key takeaway was that AI is best used to augment, and not replace, human investigators.

What's next for Sentinel AI – AI-Powered Mule Account Detection

Our long-term goal is to make Sentinel AI a production-ready platform for financial institutions to use for investigation.

Future improvements include:

Real machine learning training of financial transaction datasets. Use of Graph Neural Networks (GNNs) to detect mule account networks. The seamless integration with banking systems, enabling real-time transaction monitoring.

  • Providing Explainable Artificial Intelligence (XAI) to explain each prediction. Prioritizing investigations automatically on the basis of the level of risk. Facilitating collaborative investigations with audit trails and investigator assignment.

In the end, we can see Sentinel AI developing into an intelligent decision support tool to help financial institutions identify mule accounts faster, cut down on investigation time, and better deal with financial crime without sacrificing human oversight.

Built With

Share this project:

Updates

Submission history