👋 Hi! I am Subhodip Panda
I am a Senior Research Scientist at Fujitsu Research India, where I am a member of the Security Sciences Lab (SSL). Previously, I worked as Ph.D sciences intern at Microsoft India broadly working in the area of LLM personalization. I also served as Ph.D research intern at Adobe Research, Bangalore, where I worked on data attribution for vision language models. Prior to this, I spent a brief period as a Research Associate at Oneirix Labs, where I contributed towards development of algorithms grounded on computational statistics, with a particular emphasis on their applications in Medical AI. During my undergraduate studies, I also worked as a research intern at Laboratory for Electro-Optics Systems (LEOS) of Indian Space Research Organization (ISRO).
Recently, I have completed my Ph.D from the ECE department, Indian Institute of Science (IISc), Bangalore, where I was a member of the Representation Learning Lab, advised by Prof. Prathosh A.P. During my doctoral studies, I have had the privilege of collaborating with several distinguished researchers — Dr. Ananda Theertha Suresh, Prof. Sai Praneeth, Prof. Shubhada Agrawal, and Prof. Debabrota Basu. Prior to this, I completed my post-graduate studies in Statistics from Indian Statistical Institute (ISI), Chennai under the supervision of Prof. Sudheesh Kumar K., and my undergraduate studies in ECE from Indian Institute of Engineering Science and Technology (IIEST),Shibpur.
🔍 Research Interests
Differential Privacy, Machine Unlearning
Conformal Prediction, Calibration
Statistics, Learning Theory, Information Theory
📚 Brief Research Summary (click to expand)
My broader research interest lies in Trustworthy Machine Learning. In particular, my broad research focuses on the design and analysis of privacy- and uncertainty-aware learning algorithms. I am curious about understanding the contribution of individual data points in the learning process and developing techniques to estimate and unlearn the effect of specific data. I believe Differential Privacy offers a powerful theoretical lens into this question, and Influence Estimation (Data Attribution), Machine Unlearning provide practical tools to achieve it. I also work on topics related to Diffusion Models and Fragility issues in Bandit Algorithms. Additionally, I am very interested in Statistical Optimal Transport for statistical/deep learning problems.
📢 Recent News
Aug 2026 — 🏢 Joined Fujitsu Research India as a Senior Research Scientist.
June 2026 — 🏢 Joined Microsoft India Labs as a Ph.D sciences intern.
May 2026 — 🎓 Submitted Ph.D. thesis at the ECE department, IISc Bangalore.
April 2026 — ✈️ Attending and Presenting f-INE @ ICLR 2026 at Rio de Janeiro, Brazil.
March 2026 — 🎉 Our paper Regret Tail Characterization of Optimal Bandit Algorithms with Generic Rewards is accepted at IEEE ISIT 2026.
Feb 2026 — 🏅 Awarded as a top presenter at ACM ARCS 2026 @ IIT, Hyderabad.
Feb 2026 — 🎉 Our paper Unlearning in Diffusion models under Data Constraints: A Variational Inference Approach is accepted at TMLR 2026.
Jan 2026 — 🎉 Our paper f-INE: A Hypothesis Testing Framework for Estimating Influence under Training Randomness is accepted at ICLR 2026.
Nov 2025 — 🗣️ Invited to present PBU paper at ACM ARCS 2025 @ IIT, Hyderabad.
Aug 2025 — 📍 Attending Probabilistic and Optimization Methods workshop @ ICTS, Bangalore.
July 2025 — ✈️ Attending and Presenting our paper Concept Forgetting @ UAI 2025 at Rio de Janeiro, Brazil.
May 2025 — 🏢 Joined Adobe Research India as a Ph.D. intern.
May 2025 — 🎉 Our paper Concept Forgetting via Label Annealing is accepted at UAI 2025.
March 2025 — ✈️ Attending and Presenting PBU @ AAAI 2025 at Philadelphia, USA.
Feb 2025 — 🎉 Our paper Adapt then Unlearn: Exploiting Parameter Space Semantics for Unlearning in GANs is accepted at TMLR, 2025.
Jan 2025 — 🏅 Selected for Google Research Symposium 2025.
Dec 2024 — 🎉 Our paper Partially Blinded Unlearning: Class Unlearning for Deep Networks a Bayesian Perspective is accepted at AAAI 2025.
Oct 2024 — 🎉 Our paper FAST: Feature Aware Similarity Thresholding for Weak Unlearning in Black-Box Generative Models is accepted at IEEE Transactions on Artificial Intelligence (TAI), 2025.
Oct 2024 — 🎉 Our paper Variational Diffusion Unlearning is accepted at the NeurIPS 2024 SafeGenAI workshop.
Mar 2024 — 📄 ArXiv preprint Partially Blinded Unlearning is released.
Dec 2023 — 📄 Arxiv preprint FAST: Feature Aware Similarity Thresholding for Weak Unlearning in Black-Box Generative Models is released.
Jan 2022 — 🎓 Joined IISc Bangalore as a Ph.D. student in the ECE department.
📫 Get in Touch — Feel free to reach out to me via email if your interests align or you’d like to collaborate. I’m always happy to discuss research!