Muhammad
Taha

A complete fraud detection case study focusing on business interpretability, cost-sensitive threshold tuning, and real-world deployment considerations — not just raw accuracy metrics.

• Business-interpretable model design
• Cost-sensitive threshold tuning
• Real-world deployment considerations
• End-to-end case study pipeline

End-to-end ML project performing customer segmentation using KMeans clustering, with an interactive Streamlit dashboard for real-time segment analysis and business insight extraction.

• KMeans clustering algorithm
• Interactive Streamlit dashboard
• Real-time segment analysis
• Full ML-to-dashboard pipeline

ML-powered web app built with Streamlit to predict survival likelihood based on age, gender, and socio-economic status — demonstrating end-to-end ML model integration into a web interface.

• Classification model with interpretable output
• Streamlit web interface integration
• Socio-economic feature engineering
• Deployed ML inference pipeline

Comprehensive analysis and forecasting project for silver prices using Yahoo Finance API historical data. Includes EDA, visualization, and a 3-month forward price prediction model.

• Yahoo Finance API data pipeline
• Exploratory data analysis & visualization
• ML-based time-series forecasting
• 3-month prediction horizon

Exploratory data analysis on a real Los Angeles crime dataset. Covers data cleaning, statistical visualization, and extraction of actionable insights into crime trends and patterns.

• Real-world crime dataset EDA
• Statistical visualization & trend analysis
• Data cleaning and preprocessing
• Insight extraction from public records

Exploration of categorical data handling in Python — covering One-Hot, Label, and Ordinal encoding techniques, missing value strategies, memory optimization, and visualization best practices.

• One-Hot, Label & Ordinal encoding
• Missing value handling strategies
• Memory optimization techniques
• Visualization of categorical distributions

Beginner-friendly Jupyter Notebook demonstrating practical data import techniques in Python — reading flat files, CSV, and structured formats using built-in methods and pandas.

• Flat file and CSV ingestion
• Pandas-based data loading patterns
• Format-specific parsing techniques
• Hands-on beginner tutorial structure

Production-grade CLO (Collateralized Loan Obligation) color data processing system for traders and analysts. Features automated scheduling, rules engine, Oracle/S3 integration, and an Angular dashboard — built as a client service project.

• FastAPI backend with pluggable Oracle/Excel data sources
• AWS S3 output destination abstraction
• Rules engine, presets & cron job scheduling
• Angular 20 frontend with CLO-based column visibility

Backend system for a tattoo studio booking and inbox management platform. Handles appointment scheduling, client communication, and studio workflow management through a RESTful API architecture.

• RESTful API for booking & appointment management
• Client inbox and communication system
• Studio workflow and scheduling logic
• Secure backend with authentication

Full-stack multi-vendor e-commerce platform with admin panel, vendor dashboard, order management, and commission tracking. Production-hardened with Redis caching (85%+ hit ratio), JWT auth, and load tested at 450+ RPS sustained throughput.

• Multi-vendor system with admin & vendor dashboards
• Redis caching — 85%+ hit ratio, 7.6ms avg response
• Load tested at 450+ RPS sustained
• JWT auth, rate limiting, Helmet security headers