🛡️ TrustScan AI — Automated Fraud Intelligence & Data Processing Platform

Production-Grade AI Fraud Intelligence & Document Verification Engine
Built for multi-source data extraction, mathematical identity validation, OCR processing, and high-throughput background processing.

---

📈 Real Production Scale & Search Metrics

TrustScan is deployed live in production at trustscanai.in.
According to Google Search Console telemetry:

• 🌐 Google Search Impressions: 8,260+
• 🖱️ Organic Clicks: 617
• 📈 Click-Through Rate (CTR): 7.5% (High organic engagement)
• 🎯 Average Google Search Rank: Position #5.1 (Page 1 Google Ranking)

---

📸 Product Visual Showcase

1. Platform Landing Page & Hero Section

2. Multi-Format Fraud & Document Scanner Interface

3. Real-Time Scan Results & AI Risk Analysis Dashboard

4. Asynchronous Pipeline & Feature Architecture

---

🏗️ System & ETL Data Pipeline Architecture

TrustScan processes incoming document uploads (PDFs, images), text payloads, and external security feeds through an automated Extract - Transform - Load (ETL) pipeline:

┌────────────────────────────────┐     ┌────────────────────────────────┐     ┌────────────────────────────────┐
│       1. EXTRACT STAGE         │     │      2. TRANSFORM STAGE        │     │         3. LOAD STAGE          │
│ Multi-Source Payload Ingestion │ ──► │ Data Cleaning & Validation     │ ──► │ Dual-Database Storage &        │
│ (PDFs, Images, REST API Signals)│     │ (Python, OpenCV, OCR Rules)    │     │ Asynchronous Worker Pools      │
└────────────────────────────────┘     └────────────────────────────────┘     └────────────────────────────────┘

1. Extract Stage (Ingestion & Image Preprocessing)

• Multi-Format Extraction: Ingests raw HTTP payloads, scanned PDFs, and image streams.
• OpenCV Denoising: Applies image thresholding, grayscale conversion, and contrast alignment to prepare low-quality document scans.
• Hybrid OCR Pipeline: Combines PyMuPDF, Tesseract, and EasyOCR to extract text, bounding boxes, and structural metadata from image-only PDFs and documents.

2. Transform Stage (Data Quality, Validation & Telemetry)

• Mathematical Identity Validation:
  • Aadhaar Checksum: Validates identity numbers using the Verhoeff algorithm.
  • PAN Structural Verification: Validates registration format (Individual vs Company).
  • GSTIN / CIN Verification: Performs mathematical checksum validation against official Indian business registers.
• AI Forensics & Edit Detection: Detects metadata signatures from generative AI tools (Midjourney, DALL-E) and image manipulation tools (Photoshop, Canva).
• Telemetry Feedback Loop: Runs automated data quality routines before database writes, improving scoring accuracy by 40% and reducing false positives by 25%.

3. Load Stage & Latency Optimization (90s → <15s)

• Dual-Database Load Strategy:
  • MongoDB: Stores flexible, unstructured document payloads and OCR extractions.
  • PostgreSQL: Stores relational user data, security logs, and analytical metrics.
• Asynchronous Worker Queue Optimization:
  • Initial Bottleneck: Synchronous processing on the primary web thread took 90 seconds per request.
  • Engineering Solution: Decoupled ingestion from heavy processing by implementing asynchronous background workers and adaptive worker pool scheduling, cutting processing latency down to under 15 seconds (85% reduction).

---

🛠️ Tech Stack & Engineering Specs

Area	Technologies Used
Frontend Platform	Next.js, React.js, Tailwind CSS, Framer Motion
Backend API Server	Node.js, Express.js, WebSockets, REST APIs
Data Processing & ML	Python 3.10+, Pandas, NumPy, Scikit-learn, OpenCV, EasyOCR, PyMuPDF
Databases	PostgreSQL (Relational schema), MongoDB (Document store)
Infrastructure & DevOps	Docker, AWS (EC2/S3), Google Cloud Vision API, Git/GitHub

---

📦 Getting Started & Local Setup

Prerequisites

• Node.js: v18+
• Python: v3.10+
• Database: PostgreSQL & MongoDB (Local or Cloud Atlas)

1. Clone & Dependencies

git clone https://github.com/Dubey411/TrustScan.git
cd TrustScan

2. Frontend Setup

cd client
npm install
npm run dev

3. Backend Setup

cd server
npm install
# Configure your .env with MONGO_URI, POSTGRES_URI, and GOOGLE_CREDENTIALS
npm start

4. MLOps & Model Management

To retrain or manage the Layer-1 classifier model:

# Retrain Layer-1 Classifier
python server/scripts/train_layer1.py

# Rollback Model Version
python server/scripts/rollback.py

---

📜 License & Copyright

© 2026 TrustScan AI. All Rights Reserved.
Designed & engineered by Shubham Dubey. 🛡️💎✨