I am a Technical (Research) Associate at the Robert Bosch Centre for Cyber-Physical Systems (RBCCPS), Indian Institute of Science (IISc). My research interests include Scientific Machine Learning, Natural Language Processing, Reinforcement Learning, Robotics, Human-Robot Interaction and Multi-modal Machine Learning.

I am currently working on a project to enable richer and natural interactions with social robots. My contributions to the project include two-way remote speech communication with phoneme segmentation based lip synchronization and Wav2Vec2 architecture based speech to text conversion followed by NLP based semantic action identification and execution on social tele-robot “Asha”. Part of the work is open sourced as sonorus pypi package and imperio github repository. As a future step, in addition to tele-operation, I am also working towards enabling language (NLP) based instruction execution through reinforcement learning (RL).

Our team “Aham” (Asha) was one of the finalists in the International Conference on Social Robotics (ICSR), 2021 (Competition track) and was also among the thirty eight semi-finalists in the ANA Avatar XPRIZE competition.

I have also worked as a first co-author on the paper titled “MIMICause: Representation and automatic extraction of causal relation types from clinical notes” which was accepted in the Findings of the Association for Computational Linguistics (ACL 2022). The research pertains to defining causal annotation schema with a focus on causality types and directionality of interaction as communicated in the text between the cause and effect bio-medical concepts/entities. We created an annotated “MIMICause” corpus based on the MIMIC-III clinical notes and the National NLP Clinical Challenges (n2c2) shared task dataset; and trained baselines using BERT and Clinical-BERT based language models.

In the past at Hewlett-Packard (HP) Inc., under the guidance of Dr. Niranjan Damera-Venkata, I have worked on customer review analysis using self-attention based LSTM network and predicting printer part failures with an LSTM based architecture using multi-instance learning based hybrid loss. I received a Master’s degree from the Department of Computational and Data Sciences (CDS), Indian Institute of Science (IISc), Bangalore where I worked on information-theoretic graph characterization and generation of non-isomorphic potential drug molecular graph structures as well as shanon entropy based chemical properties prediction of hydrocarbons under the supervision of Prof. Debnath Pal and Dr. Chandan Raychaudhury. My B.Tech. thesis work under the supervision of Prof. Subrata Kumar Ghosh at IIT (ISM), Dhanbad was developing an improved mathematical model of thermal conductivity for nanofluids based on a linear conductivity gradient across the interfacial nano-layer between the nano-particles and the base fluid, and solving the resulting steady state heat conduction differential equation.