Abhishek Mann

Spearheaded development of multiple modules for “Machine Learning powered unified Video KYC chatbot framework” at IDRBT (R&D arm of Reserve Bank of India), the project is undertaken in collaboration with MeitY (Ministry of Electronics and Information Technology) and IIIT Hyderabad. Lead developer for Face Anti-Spoofing (FAS) module also known as “Liveness Detection” for determining possible spoof attempts to the face recognition pipeline; employing cutting edge deep learning algorithm(s) and advanced callback techniques to achieve state-of-the-art performance (f1-score: 0.9502, APCER-score: 0.0058) on one of the most challenging datasets OULU-NPU. Developing speech recognition pipeline employing well established auto-regressive Transformer architecture from OpenAI; for speech data of individuals pertaining to the Indian subcontinent. Advising the team for development of chatbot; intent recognition phase and the corresponding response from the chatbot. Responsible for imparting deep insight to peers about ideas from Computer Vision and NLP; including Siamese Networks and Transformers. Conducting training sessions for peers and graduate/PhD students; complete end-to-end training (including data preprocessing) of deep learning models and transfer learning using Pytorch.

Developing centralized platform to access information related to service tickets raised for client products and the steps taken by engineering team in chronological order; average ticket resolution time brought down by 30%. Took initiative for the idea of building sentiment engine for the centralized platform, performing sentiment classification for the available client emails aiding in tracking critical tickets; improvement in client satisfaction by 2%. Automating information extraction from log files of devices manufactured by client and coherent presentation of the required information; part of the team to win Spot award for innovative solution in improving efficiency of engineering team in tracking down and resolving issues.