Designed and implemented computer vision and deep learning solutions for Picsee, a photo-sharing application. Optimized models for production deployment, focusing on inference efficiency and user experience.

• HyperVerge provides identity verification and fraud detection solutions for companies in industries like BFSI, crypto, telecom, e-commerce etc. and has verified more than 1 billion identities to date.
• Contributed to information extraction workflow deployed in India, Vietnam, Indonesia & Malaysia impacting 280 million customer onboarding.
• Worked with state-of-the-art multi-modal Transformer architectures for Named Entity Recognition in ID cards and documents improving the accuracy by 30%.
• Built pipelines for preprocessing and annotation of large amounts of text and image data (0.6 B data points).
• Contributed in research and development for document quality classification tasks. Trained and deployed models to detect blur and glare in ID cards.
• Took ownership of end to end information extraction pipeline for ID cards to automate KYC for several clients in ASEAN geography.
• Trained and deployed generic OCR and Text Detection models which support multiple fonts and documents which are on par with Google OCR.

• Crea.Vision aims to automate modern living using Computer Vision. I freelanced here alongside my college studies to gain real-world industry experience.
• Deployed end-to-end system for monitoring employees in workspaces.
• Trained deep learning models for face recognition, object detection & tracking.