Shabbir Tashrifwala

Built an adversarially-trained deepfake detector robust to manipulation attacks, maintaining high accuracy on both clean and adversarially-modified videos by focusing on natural facial features.

ASL fingerspelling-to-text system using 3D pose tracking and neural networks to automatically translate hand gestures into written English with real-time inference.

Built a churn prediction system that forecasts both which customers will leave and when, generating personalized retention recommendations through daily batch processing and real-time API.