02 — PROJECTS
Selected Work
MAD-CTI: Cyber Threat Intelligence Analysis of the Dark Web Using a Multi-Agent Framework
A multi-agent LLM-based cyber threat intelligence system to scrape and analyze dark web forums for emerging malware, vulnerabilities, and exploits, improving threat detection accuracy by 14% compared to single-agent LLM approaches.
PLAiCE: Graph Neural Network for Netlist-to-Placement Prediction
A graph neural network-based placement prediction system for VLSI design, improving placement accuracy by 12% compared to traditional placement methods.
CA-Cut Revisited: A Cross-Domain Navigation Assessment and Simulated Deployment in Agricultural Robotics
Reproduction and fine-tuning of a CNN-based navigation model using targeted occlusion augmentation, achieving a 65.07% error reduction in orchard environments and integrating the model into a ROS-based robotic navigation stack for simulation testing.
EdNoteAI: AI-Powered Note Generation for Students
A web application that generates structured LaTeX lecture notes from uploaded recordings using generative AI and NLP models, implementing a scalable backend pipeline and intuitive front end to process transcripts and automate document creation.
Brain Tumor MRI Classification Model
A machine learning classification model to identify brain tumor MRI scans as glioma, meningioma, pituitary, or no tumor, achieving 94% accuracy, with a web-based interface allowing users to upload scans and view predictions.