Siddarth Asokan

RSDE @ MSRI, Bangalore | Ph.D. @ RBCCPS, IISc, Bangalore.

ProfilePic.png

Hi! I’m a Research Software Development Engineer (RSDE) at the Microsoft Research Lab (MSRI) in Bangalore, India, as part of Manik Varma’s group, and currently exploring the world of text-based generative models (with Tiny/Small/Large Language Models, as people call them these days) in the context of large-scale information retrieval (which was once the eXtreme Classification project). Previously, I was an interdisciplinary Direct-Ph.D. scholar at the Robert Bosch Center for Cyber-Physical Systems (RBCCPS) at the Indian Institute of Science, Bangalore. I worked at the Spectrum Lab in the Department of Electrical Engineering, under the supervision of Prof. Chandra Sekhar Seelamantula. Before that, I was at M.S. Ramaiah Institute of Technology (MSRIT), Bangalore, where I completed my Bachelors in Engineering in Electronics and Communications. Check out my full CV or publications. Amongst other things, I love photography and Misty, my doggo!!. For a summary of my biography, to use for public-facing documents, click here

My research interests are broadly in the area of machine learning and generative modeling. Currently, at MSR, my focus is on information retrieval (IR) models via various paradigms: Dense retrieval (DR), and generative retrieval (GR) (both autoregressive (AR) and non-autoregressive (NAR) approaches therein). Our work focuses on approaches to make IR low-latency and accurate by leveraging world knowledge, learning to transfer vocabularies across models, etc. During my Ph.D., my focus was on generative adversarial networks (GANs), and particularly on building theoretical foundations for analyzing GANs, leveraging insights from classical signal processing, and designing network architectures motivated by those findings. As my Ph.D. concluded, I (like everyone else that worked with GANs) also explored score-based diffusion and normalizing-flow models!

Throughout my Ph.D. I’ve been graciously funded by the Microsoft Research Ph.D. Fellowship in 2018, the RBCCPS Ph.D. Fellowship in 2020-2021, and 2021-2022, and the Qualcomm Innovation Fellowship in 2019, 2021, 2022 and 2023! My Ph.D. Thesis has also received the Prof. Satish Dhawan Research Award and the Indian Unit of Pattern Recognition and AI (IUPRAI) Doctoral Dissertation Awards.

Looking to join the group as a Research Fellow (A pre-Doc role that opens up around Feb-Apr) or know more about my research? Reach out to me at [Firstname].[Lastname] (at) microsoft (dot) com

News

Selected Publications

2025

  1. WWW 25
    Graph Regularized Encoder Training for Extreme Classification
    A. Mittal , S. Mohan , D. Saini , and 10 more authors
    In Companion Proceedings of the ACM on Web Conference , 2025
  2. ICML 25
    MOGIC: Metadata-infused Oracle Guidance for Improved Extreme Classification
    S. C. Prabhu , B. Singh , A. Mittal , and 11 more authors
    In Proceedings of the 42nd International Conference on Machine Learning , 2025

2024

  1. KDD 24
    Extreme Meta-Classification for Large-Scale Zero-Shot Retrieval
    S. Yadav , D. Saini , A. Buvanesh , and 6 more authors
    Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024
  2. arXiv
    CROSS-JEM: Accurate and Efficient Cross-encoders for Short-text Ranking Tasks
    B. Paliwal , D. Saini , M. Dhawan , and 6 more authors
    arXiv preprints, arXiv:2409.09795, 2024

2023

  1. CVPR 23
    Spider GANs: Leveraging friendly neighbors to accelerate GAN training
    S. Asokan , and C. S. Seelamantula
    The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023
  2. JMLR
    Euler-Lagrange Analysis of Generative Adversarial Networks
    S. Asokan , and C. S. Seelamantula
    Journal of Machine Learning Research (JMLR), 2023
  3. arXiv
    GANs Settle Scores!
    S. Asokan , N. Shetty , A. Srikanth , and 1 more author
    arXiv preprints, arXiv:2306.01654, 2023
  4. arXiv
    Data Interpolants – That’s What Discriminators in Higher-order Gradient-regularized GANs Are
    S. Asokan , and C. S. Seelamantula
    arXiv preprints, arXiv:2306.00785, 2023
  5. PhD Thesis
    On the Optimality of Generative Adversarial Networks — A Variational Perspective
    S. Asokan
    In partial fulfillment of the requirements for the Degree of Doctor of Philosophy, IISc , 2023

2020

  1. NeurIPS 20
    Teaching a GAN What Not to Learn
    S. Asokan , and C. S. Seelamantula
    In Advances in Neural Information Processing Systems (NeurIPS) , 2020