4 projects · AI/ML, Web, Computer Vision — built with real impact in mind.
2026
End-to-end voice conversion system using Whisper for speech recognition and Tacotron2/HiFi-GAN as the vocoder. Tested across 5 voice profiles with informal listening evaluations. Optimized the audio preprocessing pipeline with Librosa to significantly cut down inference time.
2026
A web application that converts spoken language in videos into sign language animations — improving accessibility for the deaf and hard-of-hearing community. Engineered a full end-to-end pipeline: audio extraction → speech-to-text → sign generation → video overlay.
2025
Real-time face landmark detection using the MediaPipe Tasks API and OpenCV in Python. Detects 468 facial landmark points from live webcam video with 30ms latency at 30 FPS. Achieved 95%+ accuracy across varied lighting conditions using the face landmarker.task model in VIDEO mode.
2024 · Hackathon
Unified ERP system for public colleges built at SIH Hackathon, eliminating fragmented data entry across departments. Delivered a fully functional prototype covering Admissions, Academics, and Real-Time Analytics — projecting 60%+ reduction in administrative overhead.