Achintya Kundu

Staff Research Scientist
IBM Research, Singapore


Email: achi [DOT] kundu [AT] gmail [DOT] com
Mobile: +65 9104 2617

Publications || Education || Achievements || CV

Academic/Professional Activities

  • August 2024: Our work on Improving Training Efficiency of LLMs Through Packing with Flash Attention is now availabe in Hugging Face.
  • March 2024: Our paper on Efficiently Distilling Large Language Models got Accepted at NAACL 2024 (Industry track).
  • February 2023: Received First Patent Application Achievement Award from IBM.
  • November 2022: Received the IBM IRL Distinguished Paper Award 2022 for the Federated Learning paper .
  • July 2022: Our Federated Learning paper won the Best Paper Award at IEEE EDGE 2022.
  • April 2022: Defended my PhD Thesis.
  • August 2021: Joined IBM Research as a Research Scientist.
  • June 2021: Submitted my PhD Thesis.
  • Research Intern, IBM Research Lab, Feb-May 2021.
    • Worked on Robust and Personalized Federated Learning.
  • Teaching Assistant, August-December 2019.
    • E0 230 (3:1) #   Computational Methods of Optimization  Course Webpage
  • Research Intern, INRIA, Paris, June-July 2018.
    • Worked on nonsmooth convex optimization with linear minimization oracle.
  • Research Intern, INRIA, Paris, September-November 2017.
    • Worked on Optimization over intersection of simple convex sets.
  • Senior Research Fellow, DST/INRIA Joint Project, IISc, Bangalore, March 2017-July 2019.
    • Worked on Learning from Big Data: First-Order methods for Kernels and submodular functions.
  • Teaching Assistant, January-April 2017.
  • Research Intern, Xerox Research Centre India (XRCI), Bangalore, July-September 2015.
    • Worked on Robust Classification and Uncertainty modelling using Copulas.
  • Teaching Assistant, August-December 2014.
  • Machine Learning Intern, Amazon, Bangalore, April-August 2014.
    • Worked on large-scale machine learning.
  • Teaching Assistant, January-April 2014.
    • E0 229 (3:1) #   Foundations of Data Science
  • Teaching Assistant, August-December 2013.
  • Teaching Assistant, August-December 2012.
  • Teaching Assistant, August-December 2011.
  • Member, Department Curriculum Committee (DCC), CSA Department, IISc, July 2010-June 2011.
  • Teaching Assistant, August-December 2010.
    • E0 230 (3:1) #   Computational Methods of Optimization  Course Webpage
  • Summer Intern, Yahoo! Labs, Bangalore, May-July 2010.
    • Worked in the field of computational advertising.
  • Design Engineer, Texas Instruments, Bangalore, July 2008-July 2009.
    • Worked on HD video compression for smart-phones.

Awards / Achievements

  • Received First Patent Application Achievement Award from IBM in 2023.
  • Received the IBM IRL Distinguished Paper Award 2022.
  • Won the Best Paper Award at IEEE EDGE 2022 for our paper on Federated Learning.
  • Recipient of Google Travel Grant for attending AISTATS 2018.
  • Won the Best Perspective Seminar Award for 2010-11 in the Dept. of CSA, IISc.
  • Won Yahoo! Key Scientific Challenges (KSC) Honorable Mention Award 2011-12.
  • Recipient of NeurIPS Travel Award 2010 sponsored by Google.
  • Received Senate Commendation from Indian Institute of Science (IISc) for outstanding academic performance in M.E.(2006-08).
  • Recipient of Prof. ISN Murthy Gold Medal 2008 for Best Master of Engineering student in the Dept. of ECE, IISc.
  • Achieved 6th rank in Graduate Aptitude Test in Engineering (GATE), 2006.
  • Ranked 17th (Engineering) and 281st (Medical) in West Bengal JEE (Joint Entrance Examination), 2002.

Publications

Book Chapter

  • P. Yu, A. Kundu, L. Wynter and S. H. Lim
    Personalized, Robust Federated Learning with Fed+
    In Book Federated Learning: A Comprehensive Overview of Methods and Applications, Springer International Publishing, 2022. Springer Link

    • Papers in AI/ML Conferences

    • A. Kundu, L. Wynter, R. D. Lee, R. K. Ganti, and M. Mishra
      Enhancing Training Efficiency Using Packing with Flash Attention
      arXiv preprint 2024. [Hugging Face Link]

    • A. Kundu, F. Lim, A. Chew, L. Wynter, P. Chong, and R. D. Lee
      Efficiently Distilling LLMs for Edge Applications
      Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL Industry Track), 2024. [arXiv] [Paper Link]

    • A. Kundu, L. Wynter, R. D. Lee and L. A. Bathen
      Transfer-Once-For-All: AI Model Optimization for Edge
      IEEE International Conference on Edge Computing and Communications (IEEE EDGE), 2023. [arXiv] [Paper Link]

    • A. Kundu*, P. Yu*, L. Wynter and S. H. Lim     (*: equal contribution)
      Robustness and Personalization in Federated Learning: A Unified Approach via Regularization
      IEEE International Conference on Edge Computing and Communications (IEEE EDGE), 2022. [arXiv] [Paper Link] [won Best Paper Award]

    • A. Kundu, F. Bach and C. Bhattacharyya
      Convex Optimization over Intersection of Simple Sets: improved Convergence Rate Guarantees via an Exact Penalty Approach.
      Conference on Artificial Intelligence and Statistics (AISTATS), 2018. [Paper Link]

    • A. Kundu, V. Tankasali, C. Bhattacharyya and A. Ben-Tal
      Efficient algorithms for learning Kernels from multiple Similarity matrices with general convex loss functions.
      Conference on Neural Information Processing Systems (NeurIPS), 2010. [Paper Link]

    Papers in Computer Systems Conferences

    • P. Pipada, A. Kundu, K. Gopinath, C. Bhattacharyya, S.Susarla and P.C. Nagesh
      LoadIQ: Learning to Identify Workload Phases from a Live Storage Trace.
      USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage), 2012. [Paper Link]

    Papers in Signal Processing Conferences

    • A. Kundu, S. Chatterjee, A.S. Murthy and T.V. Srinivas
      GMM Based Bayesian Approach to Speech Enhancement in Signal / Transform Domain.
      International Conference on Acoustic, Speech, and Signal Processing (ICASSP), 2008. [Paper Link]

    • A. Kundu, S. Chatterjee and T.V. Srinivas
      Speech Enhancement Using Intra-frame Dependency in DCT Domain.
      European Signal Processing Conference (EUSIPCO), 2008. [Paper Link]

    • A. Kundu, S. Chatterjee and T.V. Srinivas
      Subspace Based Speech Enhancement Using Gaussian Mixture Model.
      InterSpeech, 2008. [Paper Link]

    Education

    • Ph.D. ( Machine Learning )   2022   [CGPA:   7.5 / 8]
      •           Machine Learning Lab, Department of Computer Science & Automation (CSA), Indian Institute of Science (IISc), Bangalore.
                Research Supervisor: Prof. Chiranjib Bhattacharyya
                Thesis: Novel First-order Algorithms for Non-smooth Optimization Problems in Machine Learning [Thesis Link]

    • M.E. ( Signal Processing )   2008   [CGPA:   7.8 / 8]
      •           Department of Electrical Communication Engineering (ECE), Indian Institute of Science, Bangalore.
                Research Supervisor: Prof. T.V. Srinivas
                Thesis: Speech Enhancement - A Bayesian Estimation Approach using GMM

    • B.E. ( Electronics & Tele-communication )   2006   [CGPA:   9.4 / 10]
      •           Department of Electronics & Telecommunication Engineering (ETCE), Jadavpur University (JU), Kolkata.

    Research Areas

    • Large Language Models
    • Deep Learning
    • Machine Learning
    • Convex Optimization
    • Signal Processing