Hamza Alzahrani

Select a cut, stone, metal, and setting from a taxonomy of 54 ring variants, and the model renders a photorealistic finished ring in seconds. I built the entire generative core: designed the four-axis attribute taxonomy, built a synthetic data pipeline of over 1,200 images, and performed rank-4 LoRA fine-tuning of Stable Diffusion XL, achieving 97.5% attribute alignment in my evaluation. The system also includes a web frontend for end users and anyone to use. This bridges product design, generative AI, and data engineering in a single tool. The web frontend was built with a teammate.

• Stable Diffusion XL
• LoRA
• Diffusers
• PyTorch

through a four-tier auditable pipeline: motion detection, person and weapon recognition, pose estimation, and skeleton-based action classification. The system flags anomalies and builds a complete evidence chain that operators can audit in a dashboard. A five-person team project. My contributions: project planning and the software project management plan (SPMP), literature review and problem framing, and the role-based access control design for the operator dashboard.

• RT-DETR
• BoT-SORT
• RTMPose
• ST-GCN
• PyTorch