PeopleofSuperintelligence

Researcher

Harish Guruprasad

Also known as Harish Guruprasad Ramaswamy

Harish Guruprasad is an AI researcher and Assistant Professor at IIT Madras whose work spans statistical learning theory, multiclass classification, visual explanations and neural-network feature learning. After earning a B.Tech. in Electrical Engineering at College of Engineering Guindy and an M.E. in Systems Science and Automation at the Indian Institute of Science (IISc), he completed his PhD in IISc's Computer Science and Automation department under Shivani Agarwal in June 2015. Between 2014 and 2018, he coauthored papers on convex calibration dimensions for multiclass loss matrices, kernel embeddings for mixture proportion estimation and consistent multiclass classification with an abstain option. He worked as a research scientist at IBM research labs and as a postdoc at the University of Michigan before teaching at IIT Madras, where he coauthored the WACV 2020 best-paper winner Ablation-CAM along with studies of attention-network interpretability and half-space feature learning in deep networks. As of October 2026, he serves on the faculty of the Wadhwani School of Data Science and Artificial Intelligence at IIT Madras, on lien from the Department of Computer Science and Engineering.

Timeline

  1. By June 2015

    Engineering degrees and IISc doctorate

    He earns a B.Tech. in Electrical Engineering from College of Engineering Guindy, Anna University, and an M.E. in Systems Science and Automation from the Indian Institute of Science (IISc). Advised by Shivani Agarwal in IISc's Computer Science and Automation department, he completes his PhD thesis, "Design and Analysis of Consistent Algorithms for Multiclass Learning Problems", in June 2015.

  2. 2014–2016

    Calibrating multiclass losses and mixtures

    In August 2014, he and Shivani Agarwal submit "Convex Calibration Dimension for Multiclass Loss Matrices", which appears in the Journal of Machine Learning Research in 2016. At ICML in 2016, he, Clayton Scott and Ambuj Tewari publish "Mixture Proportion Estimation via Kernel Embeddings of Distributions", giving a provably correct algorithm for mixture proportion estimation with convergence rates.

  3. 2015–2018

    Multiclass classification with abstention

    With Ambuj Tewari and Shivani Agarwal, he submits "Consistent Algorithms for Multiclass Classification with a Reject Option" in May 2015. In 2018, the three coauthors publish "Consistent Algorithms for Multiclass Classification Problems with an abstain Option" in the Electronic Journal of Statistics.

  4. By 2019

    IBM, Michigan and IIT Madras teaching

    He works as a research scientist at IBM research labs and as a postdoc at the University of Michigan before joining IIT Madras. At the IIT Madras Department of Computer Science and Engineering, he teaches Probabilistic Graphical Models and Nonlinear Optimisation in 2018, and adds Advances in Theory of Deep Learning in 2019.

  5. 2020

    Ablation-CAM visual explanations at WACV

    At WACV in 2020, he and Saurabh Desai coauthor "Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-free Localization", which wins a WACV 2020 best paper award. The method uses ablation analysis to weight individual feature-map units with respect to class, creating gradient-free visual explanations for convolutional neural networks.

  6. 2022–2023

    Interpreting attention networks at ACML

    In 2022, he coauthors "On the Interpretability of Attention Networks" with Lakshmi Narayan Pandey and Rahul Vashisht for the 14th Asian Conference on Machine Learning (ACML); PMLR publishes the paper in 2023. The study defines selective dependence classification and evaluates sparsity methods for attention-model interpretability.

  7. 2024–2025

    Feature learning in neural networks

    In April 2024, he coauthors "Half-Space Feature Learning in Neural Networks" with Mahesh Lorik Yadav and Chandrashekar Lakshminarayanan, showing that deep linearly gated networks learn features through half-spaces in the input space. In 2025, he coauthors the ICLR paper "Deep Networks Learn Features From Local Discontinuities in the Label Function" with Prithaj Banerjee, Mahesh Lorik Yadav and Chandrashekar Lakshminarayanan.

  8. As of October 2026

    Assistant professor at IIT Madras

    As of October 2026, he is an Assistant Professor at IIT Madras in the Wadhwani School of Data Science and Artificial Intelligence, on lien from the Department of Computer Science and Engineering. His research interests include machine learning, learning theory and optimisation.

Career

  1. NowIIT MadrasAssistant Professor
  2. IBMResearch Scientist
  3. University of MichiganPostdoctoral Researcher

Education

  1. Anna UniversityB.Tech.
  2. Indian Institute of ScienceM.E.

From the biography.

  1. The United States employs an estimated 48% of people shaping AIIncludes IIT Madras. We estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.
  2. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes IIT Madras. We estimate where the people shaping AI studied and where their careers led. Massachusetts Institute of Technology leads with 467 former students, including 291 recorded doctorates.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes IIT Madras. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. Google has the highest combined h-index among AI organizationsIncludes IIT Madras. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Harish Guruprasad yet. Send us a free-licensed photo.

  1. Harish Guruprasad | Wadhwani School of Data Science and Artificial Intelligence wsai.iitm.ac.in. Checked 22 October 2026.
  2. Harish Guruprasad Ramaswamy - Bio sites.google.com. Checked 22 October 2026.
  3. Harish Guruprasad Ramaswamy - Research sites.google.com. Checked 22 October 2026.
  4. Convex Calibration Dimension for Multiclass Loss Matrices arxiv.org. Checked 22 October 2026.
  5. Mixture Proportion Estimation via Kernel Embeddings of Distributions proceedings.mlr.press. Checked 22 October 2026.
  6. Consistent Algorithms for Multiclass Classification with a Reject Option arxiv.org. Checked 22 October 2026.
  7. Department of Computer Science and Engineering | Indian Institute Of Technology Madras , Chennai cse.iitm.ac.in. Checked 22 October 2026.
  8. Ablation-CAM: Visual Explanations for Deep Convolutional Network via Gradient-free Localization openaccess.thecvf.com. Checked 22 October 2026.
  9. 14th Asian Conference on Machine Learning acml-conf.org. Checked 22 October 2026.
  10. On the Interpretability of Attention Networks proceedings.mlr.press. Checked 22 October 2026.
  11. Half-Space Feature Learning in Neural Networks arxiv.org. Checked 22 October 2026.