Back to projects

Fraud Free

A large-scale fraud case management system that ingests reports from multiple channels, automates lead assignment, and tracks end-to-end case resolution across internal teams and customers.

PHPLaravelMySQLRedisQueues / Cron JobsREST APIsWebhooksSpatie RBACLaravel TelescopeGPT-4 Turbo API

Overview

Fraud Free is a large-scale fraud reporting and case management system designed to handle real-world consumer fraud cases across India.

The platform aggregates fraud reports and leads from multiple channels including website forms, WhatsApp Business, Instagram DMs, and Facebook messages, and converts them into structured cases processed through a multi-stage internal workflow.

I built the entire system end-to-end, including backend architecture, integrations, workflow engine, deployment, and production maintenance.

At scale, the system handled 40,000+ users, 15,000+ fraud reports, and 800–1000 daily registrations.

Key Features

  • Multi-Channel Lead Ingestion - Unified ingestion pipeline for website forms, WhatsApp Business API, Instagram, and Facebook using webhook-based integrations.
  • Automated Lead Assignment Engine - Cron-based system that assigns leads to executives based on workload, specialization, geography, fraud type, and priority rules.
  • Case Management System - Converts qualified leads into structured cases with full lifecycle tracking from registration to resolution.
  • Customer Portal - Users can log in to track case progress, communicate with assigned agents, and view real-time updates.
  • Agent Communication Workflow - Internal communication system allowing executives to update case status and interact with customers.
  • Progressive Workflow State Machine - Strict status transition system preventing invalid flows (e.g., “registered” → “closed” directly is blocked).
  • Performance & Productivity System - Automated reporting for weekly/monthly performance, case resolution metrics, and executive productivity tracking.
  • Commission Calculation Engine - Dynamic payout calculation based on case closures, revenue contribution, and resolution time.
  • Role-Based Access Control (RBAC) - Fine-grained permissions system using Spatie to manage roles across sales, drafting, and admin teams.

System Architecture

The system operates as a workflow-driven backend where:

  • All incoming leads are normalized into a unified schema
  • Leads enter a central ingestion pipeline
  • A scheduling engine assigns work based on rules and workload balancing
  • Each lead transitions into structured cases based on business logic
  • All actions are logged for auditability and reporting

This effectively functions as a CRM + workflow orchestration system built for high-volume operations.

Key Engineering Highlights

Multi-Source Lead Aggregation

Leads were collected from multiple external systems:

  • Website forms
  • WhatsApp Business API
  • Instagram DMs
  • Facebook Messenger

All sources were unified into a single processing pipeline using webhook-based ingestion and API integrations.

Automated Lead Distribution System

A custom cron-based engine distributes leads using multiple factors:

  • Executive workload balancing
  • Fraud category specialization
  • Geographic region
  • Lead severity and value
  • Source channel prioritization

The system also includes throttling logic to prevent overloading active executives.

Workflow State Machine

A strict workflow engine enforces valid transitions across stages.

  • Each status defines allowed next states
  • Invalid transitions are blocked at the application level
  • Separate workflow paths exist for sales, drafting, and resolution teams

This ensured process integrity and prevented inconsistent case states.

AI-Assisted Drafting System

Integrated GPT-4 Turbo API into the drafting workflow.

Draft executives can generate structured complaint emails by:

  • Selecting predefined topics
  • Adding minimal contextual input
  • Allowing the system to generate complete, structured communication based on case data

This significantly reduced manual effort in drafting formal communications.

Advanced Filtering & Query System

Built a dynamic filtering engine supporting:

  • Include / exclude filters
  • Multi-value conditions
  • Historical state filtering (e.g., cases that were ever in a given status)
  • Complex nested query combinations

This powered both operational dashboards and analytics views.

Automation of Manual Operations

Previously, third-party BSP message data aggregation and reconciliation required manual processing through dedicated operational workflows.

This was fully automated into a unified ingestion pipeline, eliminating repetitive manual data handling and significantly improving processing speed and reliability across communication channels.

Observability & System Debugging

  • Laravel Telescope used for request tracing and debugging
  • Full audit logging for user and system actions
  • SMTP-based email notification system for alerts and workflows

Impact

  • Processed 15,000+ fraud reports
  • Served 40,000+ users
  • Handled 800–1000 daily registrations
  • Fully automated multi-channel lead ingestion and assignment workflows
  • Eliminated major manual operational bottlenecks in data processing
  • Improved case resolution tracking and operational visibility
  • Enabled scalable fraud case management across multiple teams

Notes

This system replaced a fragmented legacy portal that suffered from inconsistent workflows, poor data handling, and manual operational dependencies.

It was rebuilt as a production-grade workflow system in approximately 2 months, followed by continuous enhancements and scaling in production.

I was responsible for the entire system independently, including backend engineering, integrations, deployment, and production operations.