Research Profile
About Me
Building trustworthy machine learning systems that learn collaboratively while preserving privacy, security, and efficiency.
Who I Am
I am a Postdoctoral Fellow at the Indian Institute of Science (IISc), working at the intersection of federated learning, representation learning, and AI security. I investigate how distributed learning systems can be made robust against adversarial attacks — particularly in federated and parameter-efficient frameworks — and design algorithms that learn effectively from decentralized, heterogeneous data.
Research Vision
My goal is to build privacy-preserving, robust, and trustworthy distributed AI systems. I envision a future in which institutions and devices collaborate on shared models without exposing sensitive data — pursuing federated and self-supervised methods that scale gracefully while upholding rigorous standards of privacy, security, and reliability in high-stakes domains such as healthcare.
Focus
Research Areas
Federated Learning
Designing robust distributed learning algorithms that train collaboratively across decentralized data with adaptive aggregation, personalization, and privacy guarantees.
Representation Learning
Advancing self-supervised and contrastive methods that learn transferable, task-agnostic features from unlabeled and heterogeneous data across domains and modalities.
Medical AI
Building annotation-efficient, robust models for medical imaging and clinical decision support, where data scarcity, reliability, and patient safety demand extra care.
AI Security & Privacy
Analyzing and defending learning systems against gradient inversion, prompt poisoning, and other attacks that threaten data privacy and model integrity.
Current Research Questions
- 1
How can federated learning frameworks be made robust against adversarial attacks in both unimodal and multimodal settings?
- 2
How do parameter-efficient fine-tuning (PEFT) and foundation models interact with privacy and security guarantees in distributed training?
- 3
How can self-supervised representation learning deliver label-efficient medical AI without degrading under domain and data heterogeneity?
- 4
Can distributed optimization be made simultaneously private, communication-efficient, and robust to malicious participants?
- 5
How do we reconcile personalization and generalization in federated learning under realistic non-IID data distributions?
Long-Term Goals
Trustworthy distributed AI — establish frameworks whose privacy and robustness guarantees match the performance of centralized training.
Sample-efficient representation learning — develop self-supervised methods that transfer across domains and modalities with minimal supervision.
Secure collaboration at scale — enable institutions to build shared models without exposing data or weakening security.
Research to real-world deployment — bridge advanced machine learning research and dependable systems in healthcare and other sensitive domains.
Collaboration
Great research happens at the boundaries between disciplines and institutions. I actively seek partnerships with academic researchers and industry practitioners in federated learning, self-supervised learning, medical AI, and AI security — collaborations that move trustworthy distributed AI from theory to practice through joint publications, shared infrastructure, and applied deployments in privacy-sensitive domains. If your work intersects these areas, I would be glad to explore how we can build together.
Toolkit
Methodologies & Tools
Research Collaborations
Building partnerships and advancing research together
University of Lille 1 S&T
Collaboration with Dr. Tanmoy Mondal
LUT University, Sweden
Collaboration with Dr. Rajkumar Saini
IIT Kharagpur
Collaboration with Dr. Saumik Bhattacharya
NIT Durgapur
Collaboration with Dr. Sayantari Ghosh
ISI Kolkata
Active member and contributor
Future Collaborations
More research partnerships coming soon
Latest Announcements
Recent publications and project updates
Automated List of Papers for Several Research Topics
Auto-scrolling list. Hover to pause and click any topic.
Publications
Peer-reviewed research contributions to the scientific community
Journey
A timeline of my academic and professional milestones
Post-Doctoral Fellow
Indian Institute of Science • Bangalore, India
Exploring the vulnerabilities in FL Frameworks in both Unimodal and Multimodal Settings.
Under supervision of Prof. Anirban Chakraborty
Post-Doctoral Research Associate
Indian Institute of Science • Bangalore, India
Research on exploring vulnerabilities in Transformer-based FL Frameworks.
Under supervision of Prof. Anirban Chakraborty
Senior Research Assistant
Hong Kong Baptist University • Hong Kong
Conducted research in Federated learning with applications in medical image segmentation, augmented with Self-Supervised learning principles.
Under supervision of Prof. Yiu-Ming Cheung
Ph.D. in Computer Science
Indian Statistical Institute • Kolkata, India
Thesis: Self-Supervised Learning and its Applications in Medical Image Analysis
Under supervision of Prof. Umapada Pal
M.Tech (Under Dual Degree)
Indian Institute of Engineering Science and Technology • Shibpur, Howrah, India
VLSI and Microelectronics
Under supervision of Dr. Ankita Pramanik
B.Tech (Under Dual Degree)
Indian Institute of Engineering Science and Technology • Shibpur, Howrah, India
Electronics and Telecommincation Engineering
Academic Service & Leadership
Contributing to the scientific community through various roles and initiatives
1. Editorial Activities
Journal Reviewer
Conference Reviewer
2. Teaching & Mentorship
Teaching Assistant
Mentorship
3. Tutorials & Invited Talks
LLM Based Applications in Progress
Exploring and building innovative AI-powered solutions
Developing an LLM-based application for medical document analysis and diagnostic assistance using computer vision and natural language processing.
Building an intelligent system that searches for research papers and summarizes them. Creates a new github repository containing the papers in the user's profile.
Creating an LLM-based tool that helps developers with medical image analysis.
Blog: The Owl
Learn, Share, Grow - Technical insights and tutorials
Whether you’re collaborating on a feature or just keeping your work in sync, pulling changes to a non-main branch is a daily task for most developers. Here is how to do it cleanly.
Read on MediumWhen you start serving large language models locally — like MedGemma-4B running behind a FastAPI server — you quickly encounter something that most ML engineers haven’t had to think about before...
Read on MediumLarge language models are easy to run locally now. But running a model is not the same as serving it...
Read on MediumUnlike the Python package llama-cpp-python, the llama-server executable is not pre-installed anywhere. It is part of the C++ repository and must be compiled....
Read on MediumGGUF is a fully packaged, quantized model format designed specifically for inference....
Read on MediumLarge multimodal models usually demand serious hardware. A 4B parameter model in full precision occupies roughly 8GB just for weights — and that’s before accounting for activations and KV cache during generation....
Read on MediumSegmentation masks are fundamental in computer vision applications, from medical imaging to autonomous vehicles. Visualising these masks...
Read on Mediumwith calflops and torchprofile - Learn how to measure computational complexity and efficiency of your PyTorch models...
Read on MediumIn medical imaging, handling large datasets efficiently is crucial for storage and processing purposes. Neuroimaging Informatics...
Read on MediumOn Multiple Nodes using SLURM
Read on Medium10 articles available