A multimodal AI system built from scratch using PyTorch. Processes video, audio, and text inputs to detect emotional states in real-time. Deployed on AWS with 98% Match Rate.