CoreSight

An AI-powered attendance system built on CNN-based 128-dimensional facial encoding. The system performs real-time student recognition from a classroom camera feed, writes attendance records to SQLite, and exposes a FastAPI service for integration with existing academic systems. It was benchmarked across varied lighting and pose conditions, achieving 86.5% system accuracy for real-time tracking.

• 128-D facial embedding pipeline with cosine-distance matching
• Real-time inference on commodity hardware (~30 FPS)
• Published as a research paper at the ICSEAIS 2026 conference

AI Legal Risk Analyzer

An AI-powered platform automating legal contract processing, risk identification, and compliance analysis. Engineered a microservices-based architecture using Python, incorporating LLMs and OCR for automated clause extraction and legal insights. Implemented enterprise-grade security features including RBAC, encryption, and audit logging to ensure data integrity and tenant isolation.

• LLM + OCR pipeline for automated clause extraction and legal insights
• Microservices architecture with enterprise-grade RBAC and encryption
• Comprehensive audit logging and tenant isolation for data integrity

Gaming & Academic Performance Prediction

A machine learning pipeline using scikit-learn to analyze lifestyle factors affecting student grades. The project covers end-to-end workflow: data preprocessing, feature scaling, model training with Linear Regression, and cross-validation to ensure robust, generalizable predictions.

• End-to-end ML pipeline from raw data to validated predictions
• Feature scaling and preprocessing for reliable model inputs
• Linear Regression with cross-validation for robust evaluation

RepoAI

A Python-based tool for automated GitHub repository data extraction and quality analysis. It ingests any public GitHub repository and produces a structured quality report — commit cadence, contributor distribution, issue health, documentation coverage, and maintainability heuristics — combined into a single composite score with actionable insights.

• Pulls and normalizes 10+ repository signals via the GitHub API
• Weighted scoring model with explainable per-metric breakdown
• Generates a clean visual report for engineering managers