Kinjawl Bhattacharyya

Hi! I’m currently a Research Intern in the PALM Lab at the University of South Florida, working with Prof. Anshuman Chhabra on reliability for language models, focusing on abstention, epistemic uncertainty, and reducing overconfident hallucinations using data-centric methods and, more recently, RLVR.

I graduated in 2024 with a B.Tech. in Electrical Engineering from the Indian Institute of Technology(IIT) Kharagpur. Broadly, I’m interested in building learning systems that can reason and generalize well and can communicate their limits. My work spans robust/domain generalization (often via generative modeling, e.g., EBMs and diffusion/consistency models), as well as interpretability and representation learning for multimodal generation.

Previously, I worked in the Representation Learning Lab at IISc Bangalore with Prof. Prathosh A P, including projects on medical imaging robustness (with publications in Medical Image Analysis and an ICML poster). I also interned at the Vision and Learning Lab, University of Alberta, supported by the Mitacs Globalink Research Internship.

I’m applying for PhD programs (Fall 2026) in trustworthy & robust machine learning. I’m also open to research internships / RA roles (6–12 months, remote or onsite) aligned with model reliability and robust generalization. You can find more details in my CV and Google Scholar. Feel free to reach out at kinjawl03@gmail.com or via LinkedIn.

news

Aug 01, 2025

Started working as a Research Intern at the PALM Lab, University of South Florida under the supervision of Prof Anshuman Chhabra.

May 01, 2025

Our paper “LangDAug: Langevin Data Augmentation for Multi-Source Domain Generalization in Medical Image Segmentation” has been accepted as a poster at ICML 2025.

May 15, 2024

Started working as a Pre-Doctoral Fellow at the ECE Department, IISc Bangalore, under the supervision of Dr. Prathosh AP, supported by the Kotak IISc AI-ML Centre Pre-Doctoral Fellowship.

Apr 27, 2024

Successfully completed my B.Tech in Electrical Engineering from IIT Kharagpur.

Nov 13, 2023

Our paper titled Cycle consistent twin energy-based models for image-to-image translation has been accepted in Elsevier’s Medical Image Analysis(IF: 11.8) journal.

selected publications

ICML 2025

LangDAug: Langevin Data Augmentation for Multi-Source Domain Generalization in Medical Image Segmentation

Piyush Tiwary, Kinjawl Bhattacharyya, and Prathosh AP

In Forty-second International Conference on Machine Learning, 2025

Abs Code Website

Medical image segmentation models often struggle to generalize across different domains due to various reasons. Domain Generalization (DG) methods overcome this either through representation learning or data augmentation (DAug). While representation learning methods seek domain-invariant features, they often rely on ad-hoc techniques and lack formal guarantees. DAug methods, which enrich model representations through synthetic samples, have shown comparable or superior performance to representation learning approaches. We propose LangDAug, a novel Langevin Data Augmentation for multi-source domain generalization in 2D medical image segmentation. LangDAug leverages Energy-Based Models (EBMs) trained via contrastive divergence to traverse between source domains, generating intermediate samples through Langevin dynamics. Theoretical analysis shows that LangDAug induces a regularization effect, and for GLMs, it upper-bounds the Rademacher complexity by the intrinsic dimensionality of the data manifold. Through extensive experiments on Fundus segmentation and 2D MRI prostate segmentation benchmarks, we show that LangDAug outperforms state-of-the-art domain generalization methods and effectively complements existing domain-randomization approaches.
Med. Image Anal.

Cycle consistent twin energy-based models for image-to-image translation

Piyush Tiwary, Kinjawl Bhattacharyya, and Prathosh AP

Medical Image Analysis, 2024

Abs HTML Code

Domain shift refers to change of distributional characteristics between the training (source) and the testing (target) datasets of a learning task, leading to performance drop. For tasks involving medical images, domain shift may be caused because of several factors such as change in underlying imaging modalities, measuring devices and staining mechanisms. Recent approaches address this issue via generative models based on the principles of adversarial learning albeit they suffer from issues such as difficulty in training and lack of diversity. Motivated by the aforementioned observations, we adapt an alternative class of deep generative models called the Energy-Based Models (EBMs) for the task of unpaired image-to-image translation of medical images. Specifically, we propose a novel method called the Cycle Consistent Twin EBMs (CCT-EBM) which employs a pair of EBMs in the latent space of an Auto-Encoder trained on the source data. While one of the EBMs translates the source to the target domain the other does vice-versa along with a novel consistency loss, ensuring translation symmetry and coupling between the domains. We theoretically analyze the proposed method and show that our design leads to better translation between the domains with reduced langevin mixing steps. We demonstrate the efficacy of our method through detailed quantitative and qualitative experiments on image segmentation tasks on three different datasets vis-a-vis state-of-the-art methods.