I have recently joined Walmart Global Tech (earlier known as Walmart Labs) as Staff Data Scientist.
Earlier I was a Research Scientist at Parallel Computing Lab, Intel Labs, India, where my primary focus was on kernel optimization of deep learning workloads on Intel architectures (IA). For example, my code for convolution using Winograd, RNN, LSTM and GRU are available in open source libraries: LIBXSMM and Intel MKL-DNN. These libraries have been adopted in several software products including TensorFlow, Caffe, MS CNTK, Apache MXNet, Chainer, OpenVINO among others for enhanced performance on IA.
I am also interested in low-precision deep neural networks. Specifically, together with my colleagues in Intel Labs, we have developed and implemented Ternary Residual Networks which uses 8-bits for activations and 2-bits for weights (with residual edges, if required) for neural networks. I have also helped showcase the efficacy of BFLOAT16 datatype on IA.
These works have been accepted in venues such as, SuperComputing, IPDPS, ICLR, CLUSTER, and have been recognized with awards such as, ISC Best Research Poster Award (AI & ML track), Intel's Gordy Award (Intel Labs' highest award) and Divisional Recognition Award.
I have also contributed to Intel's accelerator for deep learning training as part of Intel Artificial Intelligence Products Group.
Prior to joining Intel, I received my PhD from the Department of Computer Science and Engineering, IIT Kharagpur. My research areas encompassed program analysis, formal methods and verification. I was a recipient of Senior Research Fellowship from the Department of Science and Technology, India, and TCS Research Fellowship from Tata Consultancy Services for supporting my doctoral studies. My dissertation work won Best PhD Thesis Award at VLSI Design, Best PhD Forum Paper at ISVLSI and Techno Inventor Award (PhD) from India Electronics & Semiconductor Association (IESA).
Designation: Senior Research Fellow
Project Title: Extending the scope of equivalence checking in complex embedded system design verification
Sponsoring Agency: Department of Science and Technology, Govt. of India
Duration: Jul 2009 - Oct 2012
From Pixels To Words: A Scalable Journey Of Text Information From Product Images To Retail Catalog.
Pranay Dugar, Rajesh Shreedhar Bhat, Asit Sharad Tarsode, Uddipto Dutta, Kunal Banerjee, Anirban Chatterjee, Vijay Srinivas Agneeswaran.
International Conference on Information and Knowledge Management (CIKM), November 2021, (accepted).
Exploring Alternatives to Softmax Function. Kunal Banerjee, Vishak Prasad C, Rishi Raj Gupta, Karthik Vyas, Anushree H, Biswajit Mishra.
Deep Learning Theory and Applications (DeLTA), July 2021, pp: 81-86. [arXiv]
Nominated for "Best Poster Award"
Harnessing Deep Learning via a Single Building Block.
Evangelos Georganas, Kunal Banerjee, Dhiraj Kalamkar, Sasikanth Avancha, Anand Venkat, Michael Anderson, Greg Henry, Hans Pabst, Alexander Heinecke.
International Parallel & Distributed Processing Symposium (IPDPS), May 2020, pp: 222-233. [arXiv]
(Preliminary version accepted as research poster in SuperComputing 2019.)
Reliability Evaluation of Compressed Deep Learning Models.
Brunno F. Goldstein, Sudarshan Srinivasan, Dipankar Das, Kunal Banerjee, Leandro Santiago, Victor C. Ferreira, Alexandre S. Nery, Sandip Kundu, Felipe M. G. Franca.
Latin American Symposium on Circuits and Systems (LASCAS), February 2020, pp: 1-5.