Description of your solution

Fraudlens-Aura: A Symbiotic Security Co-pilot

Our solution, Fraudlens-Aura, is a fundamental reimagining of financial security. We are moving beyond the current paradigm of reactive, alert-based fraud detection to build a proactive, symbiotic security co-pilot. This is an Agentic AI framework designed not to replace human analysts, but to augment their intuition and expertise with the scale, speed, and cognitive power of a dedicated AI team.

The core problem we address is not just financial loss, but the operational bottleneck of human-led security. Analysts are inundated with false positives, and the critical window to act against sophisticated fraud is mere seconds. Aura solves this by deploying a collaborative ecosystem of specialized AI agents that work in concert with their human counterparts.

The Aura Agentic Ecosystem

Our platform is a multi-agent system where each agent has a specialized cognitive function:

Sentinel (The Proactive Hunter): Autonomously scours network-wide data using Graph Neural Networks (GNNs) to identify the faint signals of a fraud campaign's preparation phase, neutralizing threats before significant loss occurs.

Cognito (The Strategic Mind): When a threat is hypothesized, this master agent dynamically generates a bespoke, multi-step investigation and response plan, orchestrating the other agents.

Nexus (The Human-AI Interface): This agent enables the symbiotic partnership, translating complex AI strategies into clear insights for the human analyst and learning from their feedback via Reinforcement Learning from Human Feedback (RLHF).

Aegis (The Action & Defense Core): This is the remediation agent that executes defensive actions with surgical precision, such as dynamic cooldowns or context-aware account freezes.

Scribe (The Audit & Explainability Agent): To solve the "black box" problem for regulators, this agent uses GenAI to produce a complete, narrative-style post-mortem report after every incident.

How It's Used: The Aura Command Center

An analyst using Aura logs into an intelligent, narrative-driven workspace. Instead of a sea of alerts, the dashboard features a "Hypothesis Feed" from the Sentinel agent. Clicking a hypothesis opens an "Incident Workbench," a hub for human-AI collaboration. Here, the analyst finds an Evidence Locker with all relevant data (location maps, IP risk scores, flagged SMS messages) and a Strategy Room where the Cognito agent presents its recommended plan in plain English. The analyst can interact with the Nexus agent in natural language to ask questions, validate or modify the plan, and execute it with a single click. This entire workflow, from proactive hypothesis to resolution, is designed to take minutes, not hours.

Current Status & Hackathon Goals

A key strength of our proposal is that this ambitious vision is grounded in a proven, functional platform. We are not starting from scratch.

Our Technical Foundation

• Our vision for Aura is ambitious, so it is supported by a robust, production-grade technical architecture designed for scalability, speed, and real-time responsiveness. This foundation ensures our agentic AI layer has a reliable infrastructure to operate upon. The core components include:

• A high-performance FastAPI backend: This will serve as the central nervous system for our agents, powered by a core ML model designed for high accuracy and sub-200ms latency to enable real-time decision-making.

• A real-time data pipeline: Leveraging Firebase Firestore and Firebase Auth, this pipeline provides a secure and scalable infrastructure for handling sensitive transaction data and user authentication seamlessly.

• An intelligent Admin Dashboard: Built with Streamlit, this serves as the UI for the Aura Command Center, providing the crucial interface for the symbiotic partnership between our AI agents and human analysts.

Our Hackathon Sprint (Building the Aura Layer)

Our goal for Mumbai Hacks 2025 is to build the visionary agentic security layer for global payments infrastructure. Our deliverables include:

• Developing the agentic framework using a library like AutoGen.
• Integrating Google's Gemini 1.5 Pro API to power the advanced reasoning and generative capabilities.
• Building a Proof-of-Concept GNN to demonstrate Sentinel's proactive threat-hunting.
• Implementing the new UI/UX for the Aura Command Center.
• Creating the core feedback loop for RLHF.
• A cross platform mobile application for end users.
• A comprehansive admin dashboard for visualizing/modifying the user and agentic workflow
• APIs and SDKs for direct integration of our agents into existing online payments infra of vendors.