Project: Football Analysis System using Computer Vision

Built an end-to-end system to detect, track, and analyze players, referees, and the ball from broadcast football videos using YOLOv8. Implemented multi-object tracking, jersey-color based team assignment via K-Means clustering, camera motion compensation using optical flow, and team-wise ball possession analytics.

Tech: Python, YOLOv8, OpenCV, NumPy, Scikit-learn, Roboflow. 

Project: Vocal Eyes - AI powered Smart Glasses for the visually Impaired

Designing a vision-based assistive perception system that combines Vision Transformers with Scene Graph Generation to infer rich semantic relationships in real-world environments. The framework integrates graph-to-text modeling for accurate and coherent scene narration, alongside a low-latency edge–cloud architecture that balances on-device preprocessing with cloud-based inference. Built using commercial off-the-shelf hardware, the system targets scalable deployment with an estimated ~80% cost reduction, expanding accessibility for visually impaired users.

Project: AI GENESIS - Middle East’s Largest AI Hackathon.

I built an agentic AI research copilot that ingests PDFs, web pages, and YouTube videos, converts them into a vector-based knowledge base, and uses a RAG pipeline with LLMs to provide grounded, context-aware answers for academic research.