🛡️ SentinelMarket — AI-Powered Stock Anomaly Detection

🔴 Live Demo

Protecting retail investors from market manipulation with real-time ML-powered detection

View Demo • Tech Stack • Architecture • Run Locally

---

🎯 What It Does

SentinelMarket is a production-ready, full-stack data platform that detects pump-and-dump schemes and market manipulation in the Indian stock market (NSE/BSE) using:

• 🤖 Machine Learning — Isolation Forest anomaly detection with 47 engineered features
• 📊 Real-time Data Pipelines — ETL with data warehouse, data lake, and stream processing
• 📱 Social Media Intelligence — Twitter & Telegram monitoring with FinBERT sentiment analysis
• ⚡ Live Risk Scoring — 0-100 risk scores with explainability and predictive alerts

Business Impact: Designed to protect 100+ million retail investors who lose ₹10,000+ crores annually to market manipulation

---

📸 Screenshots

Main Dashboard

Real-time market overview with live indices, feature showcase, and risk monitoring

---

Live Anomaly Feed & Stock Table

Live anomaly detection feed with sortable stock table showing risk scores

---

Analytics Dashboard

Risk distribution, market health metrics, and historical trend analysis

---

Risk Alerts

Predictive alerts with crash probability forecasting 3-7 days ahead

---

Social Intelligence

Twitter & Telegram monitoring with sentiment analysis and hype detection

---

ETL Pipelines

Data engineering dashboard showing pipeline health, runs, and warehouse stats

---

Data Quality

Data quality monitoring with completeness metrics and validation reports

---

🛠️ Tech Stack

Layer	Technologies
Frontend	Next.js 16, TypeScript, Tailwind CSS, Recharts
Backend	Python 3.11, FastAPI, SQLAlchemy, pandas
Database	PostgreSQL (Supabase), SQLite fallback
ML/AI	scikit-learn (Isolation Forest), FinBERT, 47 features
Data Engineering	ETL Pipelines, Data Lake, Data Warehouse, APScheduler
Streaming	In-memory event stream (Kafka-style architecture)
Social	Twitter API (Tweepy), Telegram API (Telethon)
Deployment	Render (Backend), Netlify (Frontend)

---

🏗️ Architecture

┌──────────────────────────────────────────────────────────────┐
│                      DATA SOURCES                             │
│   📈 Stock APIs    📱 Twitter    📮 Telegram    📰 News      │
└───────────────────────────┬──────────────────────────────────┘
                            │
           ┌────────────────┴────────────────┐
           │        ETL PIPELINES            │
           │  Extract → Transform → Load     │
           └────────────────┬────────────────┘
                            │
        ┌───────────────────┼───────────────────┐
        │                   │                   │
   ┌────▼────┐        ┌─────▼─────┐       ┌─────▼─────┐
   │Data Lake│        │Data       │       │ Stream    │
   │  (Raw)  │        │Warehouse  │       │ Processor │
   └─────────┘        └─────┬─────┘       └───────────┘
                            │
                    ┌───────▼───────┐
                    │  ML ENGINE    │
                    │ • 47 Features │
                    │ • Isolation   │
                    │   Forest      │
                    └───────┬───────┘
                            │
                    ┌───────▼───────┐
                    │ RISK SCORING  │
                    │   (0-100)     │
                    └───────┬───────┘
                            │
                    ┌───────▼───────┐
                    │  FastAPI      │
                    │  30+ Endpoints│
                    └───────┬───────┘
                            │
                    ┌───────▼───────┐
                    │  Next.js UI   │
                    │  8+ Pages     │
                    └───────────────┘

---

✨ Key Features

🔍 Anomaly Detection

• Volume Spike Detection — Z-score analysis with 85% accuracy
• Price Anomaly Detection — RSI, Bollinger Bands, momentum indicators
• ML Detection — Isolation Forest trained on 6,297 data points
• Combined Risk Score — Weighted ensemble with explainability

📊 Data Engineering

• ETL Pipelines — Modular framework with error handling & monitoring
• Data Warehouse — PostgreSQL with optimized time-series queries
• Data Lake — Gzip-compressed JSON for raw data preservation
• Stream Processing — Event-driven architecture for real-time updates
• Data Quality — Completeness metrics, validation, duplicate detection

📱 Social Intelligence

• Twitter Monitoring — Real-time sentiment with FinBERT
• Telegram Channels — Pump signal detection
• Hype Score — 0-100 coordination detection

🚨 Alerts & Predictions

• Risk Alerts — HIGH/EXTREME risk notifications
• Crash Prediction — 3-7 day ahead probability forecasting
• Pattern Matching — Historical scam comparison

---

🚀 Quick Start

# Clone
git clone https://github.com/umangkumarchaudhary/SentinelMarket-Backend.git
cd SentinelMarket-Backend

# Backend
cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt
python -m uvicorn main:app --host 127.0.0.1 --port 8000 --reload

# Frontend (new terminal)
cd frontend
npm install
npm run dev

Access:

• Frontend: http://localhost:3000
• API Docs: http://localhost:8000/docs

---

📈 Performance

Metric	Value
API Response Time	<500ms avg
Detection Accuracy	~90% combined
False Positive Rate	15-20%
Stocks Analyzed/Hour	1000+
Data Quality	>95% valid ratio

---

🎯 Skills Demonstrated

This project showcases expertise in:

Area	Skills
Data Engineering	ETL Pipelines, Data Warehouse, Data Lake, Stream Processing, Data Quality
Machine Learning	Feature Engineering (47 features), Anomaly Detection, Model Deployment
Backend	FastAPI, REST APIs, PostgreSQL, SQLAlchemy, Error Handling
Frontend	Next.js, TypeScript, Responsive Design, Real-time Updates
DevOps	Render, Netlify, Docker, CI/CD
NLP	FinBERT, Sentiment Analysis, Social Media Mining

---

👤 Author

Umang Kumar Chaudhary

Building enterprise-grade data platforms and AI systems

---

📄 License

MIT License — See LICENSE for details.

---

⭐ Star this repo if you find it useful! ⭐

Built with ❤️ for protecting retail investors