Large-scale academic event management system developed for Tec de Monterrey's HiTec conference. The platform proved its robustness by handling 600+ concurrent users during registration peaks with zero downtime. I implemented a CSV bulk registration system capable of managing 1,500+ activity slots, along with a streamlined authentication method that reduced access time to just 3 seconds using student ID validation. The architecture featured: intelligent scheduling to prevent overlaps (limiting to 1 activity per hour), real-time updates via Supabase, and optimized Vercel deployment ensuring sub-2-second response times even under peak load.

AI model achieving 94.7% precision in classifying 40 car part types. Used ResNet50 transfer learning with custom dense layers (512-128 neurons) and dropout regularization. Optimized via early stopping, reducing overfitting by 23% from initial architecture.