I built an AI-based intrusion detection system that captures live TCP/UDP traffic and scores flows with machine-learning models. When I tested it with live attack traffic, I found the models missed floods outside their training distribution and added a model-independent rule that caught a real device's flood within one second.
I led the engineering of a three-person team for this project. I created reproducible evaluations using a held-out CICIDS2017 split and raw-pcap replay through the live pipeline, and diagnosed and fixed three feature-extractor mismatches that had reduced recall.
As a Web Development & Database Intern at Bureau of Agriculture Information KP, I built responsive intranet pages with HTML, CSS, JavaScript, and Bootstrap for use across phones, tablets, and desktops.
On my Underwater Trash Detection with YOLOv8 project, I fine-tuned a segmentation model on the TrashCan dataset and served detections through a Flask app. I also compared four datasets and added marine-life and ROV classes to reduce false detections. I completed the Google Cybersecurity Professional Certificate, and CompTIA Security+ is in progress.
