Debarpan Bhattacharya

I am a PhD student (starting January, 2023) at Indian Institute of Science (IISc), Bangalore, India. I was an MTech(Research) student at IISc from 2020 to 2022. Before joining at IISc, I spent amazing four years at Jadavpur University, Kolkata, while pursuing my bachelor's in Electrical Engineering.

At IISc, I am associated with LEAP lab while pursuing my PhD with Prof. Sriram Ganapathy. My research interest lies at the broad domain of reliable AI. I have also been a primary member in project COSWARA, one of the prominent responses of IISc to limit the spread of COVID-19. Check out our publications for more details.

During my bachelor's, I have worked on multiple signal processing based projects with sensing and healthcare applications. I was fortunate to receive close mentorship from Prof. Sugata Munshi, Prof. Biswajit Bhattacharyya and Prof. Sudip Misra.

LinkedIn / Google Scholar / Twitter / Github

Research

I'm broadly interested in reliable AI including explainable and interpretable AI methods, uncertainty estimation, trust assessment, and reasoning for LLMs and multimodal LLMs. Check out my publications for further details.

Contact

If you are working on similar projects, and want to have a chat, feel free to drop a mail at debarpanb[at]iisc[dot]ac[dot]in / debarpanbhatta123[at]gmail[dot]com .

Resume

download [last updated: 16/06/2026]

Recent Updates

Jun 2026: Attended Academic Summit at Microsoft Reseach, Bangalore.

Aug 2025: 1 paper accepted in EMNLP (Findings) 2025.

May 2025: 1 paper accepted in INTERSPEECH 2025.

May 2025: Joined IBM Research Lab (IRL) as a Summer Reseach Intern.

Dec 2024: 1 paper accepted in ICASSP 2025.

Sep 2024: Journal paper accepted in IEEE-JSTSP.

May 2024: Joined IBM Research Lab (IRL) as a Summer Reseach Intern.

July 2024: Selected in TCS Research Scholar (TCS-RSP), 2024 (among 3 in IISc).

June 2023: Selected in Qualcomm Innovation Fellowship, 2023.

June 2023: "Coswara: A respiratory sounds and symptoms dataset for remote screening of SARS-CoV-2 infection" is published with open access in Nature Scientific Data.

Jan 2023: Journal paper "Coswara: A respiratory sounds and symptoms dataset for remote screening of SARS-CoV-2 infection" is accepted in Nature Scientific Data.

Jan 2023: Submitted MTech(Research) thesis and rejoined to PhD at LEAP Lab, IISc.

Sep 2022: Selected for Doctoral Symposium and Demonstrations tracks in AI-ML Systems 2022 to be held in Bangalore.

June 2022: 2 full papers and 1 Show and Tell paper accepted in INTERSPEECH 2022 to be held in South Korea.

Jan 2022: 1 paper accepted in ICASSP 2022 to be held in Singapore.

Oct 2020: Started pursuing MTech(Research) at IISC Bangalore.

Achievements

TCS Research Scholar: One of the few in India, including 3 from IISc, to receive the fellowship. 2024

Qualcomm Innovation Fellowship: One of the 12 candidates in India to receive QIF. 2023

Best Paper Award: For the paper [link] at IEEE INDICON 2021. 2021

B N Paul Memorial Silver Medal: Academic excellence in B.E. at Jadavpur University. 2020

ISCA Travel Grant: Paper presentation at INTERSPEECH 2022 in South Korea. 2022

Research Week with Google: Selected for the one-week event by Google Research. 2022

Summer Research Fellow, Indian Academy of Sciences: One of 110 selected in India. 2019

Graduate Aptitude Test in Engineering: Ranked 195 out of 93,526 candidates. 2020

Publications

	Benchmarking and Confidence Evaluation of LALMs For Temporal Reasoning Debarpan Bhattacharya, Apoorva Kulkarni, Sriram Ganapathy INTERSPEECH, 2025 paper / code This paper introduces TREA, a temporal reasoning evaluation dataset for large audio language models. We benchmark open-source LALMs on audio event ordering, counting, and duration reasoning, and show that current models remain far behind human performance. We further propose an uncertainty metric based on semantically equivalent perturbations and show that accuracy and uncertainty are not necessarily correlated.
	FESTA: Functionally Equivalent Sampling for Trust Assessment of Multimodal LLMs Debarpan Bhattacharya, Apoorva Kulkarni, Sriram Ganapathy Findings of EMNLP, 2025 paper / code This paper proposes FESTA, a black-box uncertainty estimation method for trust assessment of multimodal LLMs. FESTA generates functionally equivalent and functionally complementary input samples to probe model consistency and sensitivity, enabling unsupervised selective prediction for visual and audio reasoning tasks. The method improves misprediction detection over entropy-based uncertainty measures.
	Towards Unbiased Evaluation of Time-series Anomaly Detector Debarpan Bhattacharya, Sumanta Mukherjee, Chandramouli Kamanchi, Vijay Ekambaram, Arindam Jati, Pankaj Dayama ICASSP, 2024 paper This paper studies the evaluation bias in time-series anomaly detection caused by commonly used point-adjustment protocols. We propose Balanced Point Adjustment, an alternative evaluation protocol that addresses over-estimation in detector performance and provides a fairer assessment of time-series anomaly detectors through axiomatic analysis.
	[Impact Factor: 13.7] Gradient-free Post-hoc Explainability Using Distillation Aided Learnable Approach Debarpan Bhattacharya, Amir H. Poorjam, Deepak Mittal, Mohit Singh, Sriram Ganapathy IEEE Journal of Selected Topics in Signal Processing, 2025 paper This paper proposes Distillation Aided Explainability, a gradient-free post-hoc explanation framework for black-box models. The method learns saliency masks using input-output access to the target model and a local distillation network, enabling model-agnostic explanations across modalities such as image and audio classification.
	[Impact Factor: 11.5] Coswara: A respiratory sounds and symptoms dataset for remote screening of SARS-CoV-2 infection Debarpan Bhattacharya, Neeraj Kumar Sharma, Debottam Dutta, Srikanth Raj Chetupalli, Pravin Mote, Sriram Ganapathy, C. Chandrakiran, Sahiti Nori, K. K. Suhail, Sadhana Gonuguntla, Murali Alagesan Nature Scientific Data, 2023 paper This paper presents the Coswara dataset, a dataset containing diverse set of respiratory sounds and rich meta-data, recorded between April-2020 and February-2022 from 2635 individuals (1819 SARS-CoV-2 negative, 674 positive, and 142 recovered subjects). The respiratory sounds contained nine sound categories associated with variants of breathing, cough and speech. A rich set of metadata is also present.
	Analyzing the impact of SARS-CoV-2 variants on respiratory sound signals Debarpan Bhattacharya, Debottam Dutta, Neeraj Kumar Sharma, Srikanth Raj Chetupalli, Pravin Mote, Sriram Ganapathy, Chandrakiran C, Sahiti Nori, Suhail K K, Sadhana Gonuguntla, Murali Alagesan INTERSPEECH, 2022 paper We analyze the Coswara dataset which is collected from three subject pools, namely, i) healthy, ii) COVID-19 subjects recorded during the delta variant dominant period, and iii) data from COVID-19 subjects recorded during the omicron surge. Our findings suggest that multiple sound categories, such as cough, breathing, and speech, indicate significant acoustic feature differences when comparing COVID-19 subjects with omicron and delta variants.
	Coswara: A website application enabling COVID-19 screening by analysing respiratory sound samples and health symptoms Debarpan Bhattacharya, Debottam Dutta, Neeraj Kumar Sharma, Srikanth Raj Chetupalli, Pravin Mote, Sriram Ganapathy, Chandrakiran C, Sahiti Nori, Suhail K K, Sadhana Gonuguntla, Murali Alagesan Show and Tell, INTERSPEECH, 2022 short paper / video A user using this service can log into a website using any device connected to the internet, provide there current health symptom information and record few sound sampled corresponding to breathing, cough, and speech. Within a minute of analysis of this information on a cloud server the website tool will output a COVID-19 probability score to the user.
	[Oral presentation] Acoustic Representation Learning on Breathing and Speech Signals for COVID-19 Detection Debottam Dutta, Debarpan Bhattacharya, Sriram Ganapathy, Amir Hossein Poorjam, Deepak Mittal, Maneesh Singh INTERSPEECH, 2022 paper / code We describe an approach for representation learning of audio signals for the task of COVID-19 detection. The raw audio samples are processed with a bank of 1-D convolutional filters that are parameterized as cosine modulated Gaussian functions. The relevance weighting emphasizes the key regions of the time-frequency decomposition of the filtered output. The subsequent layers of the model consist of a recurrent architecture.
	Svadhyaya system for the Second Diagnosing COVID-19 using Acoustics Challenge 2021 Deepak Mittal, Amir H Poorjam, Debottam Dutta, Debarpan Bhattacharya, Zemin Yu, Sriram Ganapathy, Maneesh Singh Pre-print, 2022 arxiv This report describes the system of Team Svadhyaya used for detecting COVID-19 positives using three different acoustic modalities, namely speech, breathing, and cough in the second DiCOVA challenge. The system reached the blind test AUCs of 86.41, 77.60, and 84.55, in the breathing, cough, and speech tracks, respectively, and the AUC of 85.37 in the fusion of these three tracks.
	The Second Dicova Challenge: Dataset and Performance Analysis for Diagnosis of Covid-19 Using Acoustics Neeraj Kumar Sharma, Srikanth Raj Chetupalli, Debarpan Bhattacharya, Debottam Dutta, Pravin Mote, Sriram Ganapathy ICASSP, 2022 paper / arxiv / slides The Second Diagnosis of COVID-19 using Acoustics (DiCOVA) Challenge aimed at accelerating the research in acoustics based detection of COVID-19. In this paper, we present an overview of the challenge, the rationale for the data collection and the baseline system. Further, a performance analysis for the systems submitted by the 21 participating teams in the leaderboard is also presented.
	[Best Paper Award] Huffman Coding based ECG Processing For Compression-Distortion Tradeoff Debarpan Bhattacharya IEEE INDICON, 2021 paper In our proposed approach, a noble huffman coding based compression technique for ECG signal is proposed. It offers high compression ratio (CR) with minimal percentage mean square difference (PRD). Additionally, the algorithm offers flexibility to decrease PRD value at the cost of CR, obtained by tuning a single parameter.
	Pulse Train Modulation And ANN Based Temperature Sensor With Semi-automatic Calibration Pranabendra Prasad Chandra, Debarpan Bhattacharya, Biswajit Bhattacharyya, Sugata Munshi IEEE INDICON, 2021 paper ANN based temperature measurement system involving thermocouples with provision of semi-automatic calibration. The proposed measurement system outperforms all the thermocouple-based temperature measurement techniques proposed in recent literature, in terms of accuracy, reliability, cost, and potential for embedded system application.
	IDeA: IoT-based autonomous aerial demarcation and path planning for precision agriculture with UAVs Debarpan Bhattacharya, Sudip Misra, Nidhi Pathak, Anandarup Mukherjee ACM Transactions on Internet of Things, 2020 paper We proposed an autonomous and onboard image-based agricultural land demarcation and path-planning system—IDeA (IoT-Based Autonomous Aerial Demarcation and Path Planning for Precision Agriculture) with Unmanned Aerial Vehicles (UAVs).
	Optimized Thermocouple Temperature Sensor using 555 Timer and ANN Based Linearization Debarpan Bhattacharya, Pranabendra Prasad Chandra, Biswajit Bhattacharyya, Sugata Munshi IEEE CALCON, 2020 paper Two-stage linearization scheme for thermocouple based temperature sensing. The first stage involves IC555 based astable multivibrator and the second stage is fully connected neural network based regression.