Sai Minnal

EventSnap is a full-stack event photo management platform that enables users to upload, organize, and discover event photos using facial recognition. Built with Python, FastAPI, JavaScript, and Face Recognition, it automatically identifies photos containing a user and creates personalized photo collections.

AI-powered ordering assistant for restaurants. Customers place orders in natural language via messaging systems, parsed by Llama 3.3 70B LLM into structured intents. Includes a real-time staff dashboard for order tracking and menu management. Built with Python, FastAPI, SQLAlchemy, and SQLite.

This project builds a modded-nanoGPT–style training harness for Vision Transformers (ViTs). Instead of optimizing only for final accuracy, we optimize for time-to-target accuracy — i.e., how fast (in wall-clock seconds) a model can reach 85% accuracy on a standard vision benchmark such as CIFAR-10.