Nivedhaa 'Nia'
Naresh Kumar

The first AI-powered roommate matching platform for college students. Built with .edu verification, lifestyle compatibility matching, a swipe deck, and realtime messaging between matches.

Machine learning pipeline to detect DDoS attacks in real network traffic using the CICIDS2017 dataset. Trained a Random Forest classifier achieving 99.99% accuracy and applied UMAP dimensionality reduction to visualize attack vs. normal traffic clusters in 3D.

End-to-end CNN/Transformer pipeline for endometrial cancer detection from ultrasound images. Achieved 89% classification accuracy using contrastive learning and custom feature engineering.

Automated data pipeline to scrape and structure financial metrics, integrating live market data via yfinance. Rule-based valuation framework using P/E ratios, revenue growth, and earnings trends.

Full-stack React application built for a Georgia Tech behavioral study exploring how physical effort influences decision-making, enabling real-time experimental tasks and structured participant data collection.

Trained a modified ResNet50 classifier with AdamW + cosine LR scheduling. Applied UMAP and distance-based outlier detection to flag noisy images and visualize dataset structure.