Contextualized Heterogeneous Graph Neural Network for extractive document summarization (AISP 2024). Treats documents as word–sentence heterogeneous graphs with TF-IDF edges, multi-kernel CNN n-grams, BiLSTM encoding, and GAT refinement with iterative message passing.

Multilingual text-processing toolkit with end-to-end preprocessing in Jupyter. Features normalization, tokenization, lemmatization, stopwords removal, spell checking, and word-cloud visualization for both English and Persian.

Hands-on scikit-learn implementations covering supervised & unsupervised algorithms with real datasets. Features SVM, Logistic Regression, KNN, Decision Trees, K-Means, Hierarchical clustering, and DBSCAN.

End-to-end recommender systems with Collaborative Filtering, Content-Based, and Hybrid approaches. Uses MovieLens (34k movies, 4M+ ratings) and Amazon reviews with comprehensive evaluation metrics.

Retail feedback mining over 82,791 clothing-store reviews to extract insights on sizing, fit, quality, and satisfaction. Features comprehensive EDA, data quality audits, and business-facing visualizations.

Full-stack MERN e-commerce application delivering a polished buyer journey and operational admin back office. Features faceted product browsing, Stripe checkout, order tracking, inventory CRUD with Cloudinary uploads, and comprehensive admin dashboard with sales analytics.

Real-time multi-client platform with React UI and Node.js/Express backend over Socket.IO, persisting to MS SQL Server. Supports multi-tenant data entry, live sync, in-app chat, and server-side analytics engine.

PyQt6 desktop app for FIFA player management on SQLite with MVC structure. Features sortable table view, multi-criteria search & filters, rating slider, and polished UX.

VHDL coprocessor for 3×3 matrix multiplication with column-by-column MAC pipeline. Features Register File, Processing Unit with multiply-accumulate, FSM Control Unit, and minimal ISA. Completes a full 3×3 matrix multiply in ~10 clock cycles.