Case studies that connect data, decisions, and delivery

Built a supervised learning pipeline to predict passenger satisfaction from operational and service quality indicators.

• Performed EDA and feature cleanup for noisy customer and flight attributes.
• Benchmarked multiple models and selected Random Forest for strongest generalization.
• Prioritized model interpretability to align with service-improvement decisions.

Designed an exploratory analytics app for player performance, squad logic, and comparative decision-making.

• Combined multi-year FIFA records into a query-ready analysis dataset.
• Built a 4-3-3 best XI generator with role-aware comparisons.
• Exposed scatter and composition views for practical team-building analysis.

Evaluated LLM behavior and human judgment patterns for hate speech identification in mixed Bangla-English text.

• Handled 1,800+ missing values and inconsistent fields before modeling.
• Compared GPT, Gemini, and Deepseek outputs with human-aligned reasoning checks.
• Contributed to conference publication and reproducibility notes.

Built a one-link survey distribution workflow to remove manual assignment complexity and prevent response imbalance.

• Used Google Sheets as a lightweight control layer and counter system.
• Implemented assignment logic with Google Apps Script and web app entry point.
• Tracked response distribution and prevented over-assignment per form target.

Developed a real-time crime reporting and safety visualization platform with emergency alert management and community safety features.

• Designed modular backend APIs for authentication, reporting, file uploads, and live updates using FastAPI.
• Built interactive map interface using Leaflet.js with crime markers, heatmaps, and GPS detection capabilities.
• Implemented emergency features including panic button, push notifications, and anonymous reporting functionality.
• Developed responsive dark/light UI with optimized usability, accessibility standards, and Tailwind CSS.
• Led backend architecture and collaborated effectively within a four-member development team.

Built and launched a production-ready full-stack e-commerce platform with complete end-to-end flows for product browsing, order placement, tracking, and admin operations.

• Developed modular backend APIs for products, orders, reviews, and admin authentication, enabling maintainable feature delivery and faster iteration cycles.
• Implemented database-driven architecture with MongoDB Atlas, replacing static data and supporting scalable CRUD operations, advanced filtering, and order-status workflows.
• Delivered a complete admin panel for product management, order processing, and review moderation, improving operational control and eliminating manual handling.
• Integrated real-time and notification workflows using Pusher, Nodemailer/Resend, and Telegram, improving response speed for new orders and status updates.
• Shipped and maintained the platform on a completely free-tier stack ($0 cost): MongoDB Atlas for database, Vercel for hosting and deployment, and GitHub for version control and storage.
• Architected and deployed with Next.js 14 and TypeScript, ensuring type safety, scalability, and modern development practices.