Content Refresh Risk Screening using Machine Learning

Developed a machine learning solution to identify web pages at risk of declining performance using a dataset of 101,362 pages across 44 anonymized clients. Compared Logistic Regression and Random Forest models and evaluated performance using ROC-AUC, F1 score, confusion matrices, and feature importance.

University Course Enrollment Monitor

Developed an automated web monitoring tool to track university course enrollment availability and detect changes in course status. Implemented automated monitoring, data extraction, and notification logic with a focus on reliability and practical automation.

DecisAi - AI Bidding Engine

DecisAI is an AI-powered bidding and procurement decision engine built for analyzing RFPs, RFQs, and tenders. It processes large, multi-page procurement documents, extracts requirements, deadlines, evaluation criteria, financial information, and proposal questions, then matches those requirements against a company’s capability and project-history dataset using Retrieval-Augmented Generation (RAG) with SentenceTransformers and ChromaDB. The system performs compliance analysis, identifies gaps and risks, generates evidence-backed proposal responses using local Ollama/Qwen, validates outputs to reduce hallucinations, and combines compliance, budget, historical performance, and competitive factors to produce a GO, CONDITIONAL GO, HIGH RISK, or NO-GO recommendation. It also supports proposal export to JSON, DOCX, and PDF and keeps the final decision human-confirmed.

Tic-Tac-Toe AI

Developed a Python-based Tic-Tac-Toe desktop game featuring an AI opponent with multiple difficulty levels. Implemented game-tree search, Alpha-Beta pruning, game-state evaluation, win detection, score tracking, theme switching, and a Tkinter-based user interface, with separate modules for game logic and UI.

Spotter Freight Rate Prediction

Developed a machine learning solution to predict freight rates using 48,000 labeled loads. Performed data analysis and preprocessing, engineered and evaluated features, implemented time-based validation to prevent data leakage, and compared Ridge Regression with tree-based models. Selected a standardized Ridge Regression model achieving a mean MAE of USD 141.29 across temporal validation folds, then generated predictions for 12,000 unseen loads.

Game Hosting Platform

Developed a web-based game hosting platform where developers can publish games and users can discover and play browser-based games. The project combines web development, database integration, search and filtering functionality, and a planned AI-based game generation workflow.

Connect Four AI

Developed a playable Connect Four desktop game in Python and Pygame with an AI opponent using the Minimax algorithm, Alpha-Beta pruning, and heuristic evaluation. Implemented the game logic, AI decision-making, difficulty levels, win/loss detection, scoreboard, and interactive user interface.

Hostel Management System

Developed a console-based Hostel Management System in C++ demonstrating object-oriented software design. Implemented student registration, automatic room allocation, fee calculation, payment receipts, administrator workflows, input validation, and file-based data persistence using classes, inheritance, polymorphism, abstraction, encapsulation, and templates.