MerchantPulse AI is an enterprise-grade analytics platform that simulates the internal data infrastructure of a digital payments company like PhonePe, Razorpay, or Google Pay.
It demonstrates the complete data lifecycle — from database design and synthetic data generation, through SQL analytics and machine learning, to interactive dashboards and a Streamlit web application.
This is not a college project. It's built to the standards expected at companies like PhonePe, ZS Associates, American Express, Deloitte, PwC, and Amazon.
Source Systems → Data Ingestion → Storage → Processing → Analytics → ML Models → Dashboards
| Layer | Technology | Purpose |
|---|---|---|
| Database | SQLite (PostgreSQL-compatible) | 12-table star schema with 2M+ transactions |
| Data Pipeline | Python, Pandas, NumPy | Cleaning, validation, feature engineering |
| SQL Analytics | 50+ production queries | Revenue, merchant, customer, fraud, geo analytics |
| Machine Learning | XGBoost, LightGBM, Prophet, K-Means | Churn, fraud, segmentation, forecasting |
| Dashboard | Power BI | 6-page executive BI dashboard |
| Web App | Streamlit | 6-page interactive analytics application |
📐 Full Architecture Document →
MerchantPulse-AI/
├── config/ # Central configuration & settings
├── database/ # Schema DDL & synthetic data generator
├── sql/ # 50+ business analytics queries
├── notebooks/ # Jupyter analysis notebooks (EDA, ML)
├── src/ # Production Python modules
│ ├── models/ # ML model implementations
│ └── utils/ # Database connector & helpers
├── streamlit_app/ # Interactive web application
├── powerbi/ # Power BI dashboard & screenshots
├── data/ # Raw → Processed → Features pipeline
├── models/ # Saved ML model artifacts
├── reports/ # Business insights & analysis
├── docs/ # Architecture, ERD, data dictionary
└── tests/ # Data quality & model tests
- Python 3.10+
- Power BI Desktop (for
.pbixdashboard)
# Clone the repository
git clone https://github.com/Harikas06/MerchantPulse-AI.git
cd MerchantPulse-AI
# Create virtual environment
python -m venv venv
venv\Scripts\activate # Windows
# source venv/bin/activate # macOS/Linux
# Install dependencies
pip install -r requirements.txt
# Launch Streamlit app
streamlit run streamlit_app/app.py- Connect your GitHub account to share.streamlit.io.
- Select repository:
Harikas06/MerchantPulse-AI(Branch:main). - Set Main file path:
streamlit_app/app.py. - Click Deploy! — The application deploys zero-config ready out of the box.
- 12 tables — 6 dimension, 4 fact, 2 aggregate
- 2M+ transactions across 5,000 merchants and 50,000 customers
- Covers Jan 2023 – Jun 2025 (2.5 years)
- 50+ business queries organized by: Revenue, Merchant, Customer, Fraud, Geographic, Operational
- Uses CTEs, window functions, cohort analysis, funnel analysis
| Model | Algorithm | Business Use |
|---|---|---|
| Customer Churn | XGBoost + LightGBM | Identify at-risk customers for retention |
| Fraud Detection | Isolation Forest | Flag suspicious transactions |
| Segmentation | K-Means + RFM | Target marketing campaigns |
| Forecasting | Prophet | Revenue planning & projections |
- Executive Summary, Merchant Deep-Dive, Customer Insights
- Geographic View, Fraud & Risk, Payment Mix
- Real-time KPIs with trend indicators
- Interactive filters (date, city, category, payment method)
- ML model playground for live predictions
| Pillar | Key Metrics |
|---|---|
| 💰 Revenue | GMV, Take Rate, ARPU, MoM Growth |
| 🏪 Merchant | Active Count, Activation Rate, Churn Rate, Health Score |
| 👤 Customer | MAU, CLV, Frequency, Cohort Retention |
| ⚡ Operations | Success Rate, Avg Ticket Size, Settlement Cycle |
| 🛡️ Risk | Fraud Rate, Detection Rate, Dispute Rate |
| Category | Technologies |
|---|---|
| Language | Python 3.10+, SQL |
| Database | SQLite (PostgreSQL-compatible schema) |
| Data | Pandas, NumPy, SciPy, Faker |
| Visualization | Matplotlib, Seaborn, Plotly |
| ML | Scikit-learn, XGBoost, LightGBM, Prophet, SHAP |
| Web App | Streamlit |
| BI | Power BI Desktop |
| Quality | Great Expectations |
- Architecture Document — Full system design
- Data Dictionary — Column-level documentation
- Business Insights Report — Key findings
This project is licensed under the MIT License.
Built as a portfolio project demonstrating enterprise-level data analytics skills for roles at digital payment companies, consulting firms, and technology organizations.
Skills Demonstrated: Database Design · SQL Analytics · Python Data Engineering · EDA · Machine Learning · Power BI · Streamlit · Business Communication