Saurabh Dash

About

I'm a Member of Technical Staff at CohereAI, where I work on improving the efficiency of very Large Language Models.
I also collaborate with Sara Hooker and others at Cohere For AI on Fundamental AI Research.

Before this, I was a PhD student at Georgia Tech (GT), advised by Dr. Saibal Mukhopadhyay.

My research interests lie broadly in building intelligent systems that are computationally efficient.

In the Summer of 2021, I was a Machine Learning research intern in the MIND team at Apple, supervised by Dr. Hema Koppula.

I was awarded the ECE fellowship at Georgia Tech during my admission into the PhD program. During my undergraduate studies at IIT Kharagpur, I received the Order of Merit for distinguished performance in the field of technology. My undergraduate thesis was also nominated for the Best Thesis Award.

Apart from my academics and personal research projects, I have helped inculcate a thriving research culture at IIT Kharagpur and share my passion for technology through the Kharagpur RoboSoccer Students’ Group – a diverse mix of freshman to senior undergraduate students from various departments working together to build robots that play soccer autonomously. As the head of the group, I was fortunate to lead the first ever Indian contingent to qualify for RoboCup SSL to Japan.

I have served as a reviewer for various machine learning conferences: NeurIPS, ICML, ICLR, TMLR, CVPR, ICCV...

Reach me at: hello.saurabhdash[at]gmail.com.

Publications

Select Publications:

How Does Quantization Affect Multilingual LLMs?.
Kelly Marchisio, Saurabh Dash, Hongyu Chen, Dennis Aumiller, Ahmet Üstün, Sara Hooker, Sebastian Ruder.
Arxiv preprint.
Aya 23: Open Weight Releases to Further Multilingual Progress.
Viraat Aryabumi, John Dang, Dwarak Talupuru, Saurabh Dash, David Cairuz, Hangyu Lin, Bharat Venkitesh, Madeline Smith, Kelly Marchisio, Sebastian Ruder, Acyr Locatelli, Julia Kreutzer, Nick Frosst, Phil Blunsom, Marzieh Fadaee, Ahmet Üstün, Sara Hooker.
Arxiv preprint.
Outliers and Calibration Sets have Diminishing Effect on Quantization of Modern LLMs.
Davide Paglieri, Saurabh Dash, Tim Rocktäschel, Jack Parker-Holder.
Arxiv preprint.
Intriguing Properties of Quantization at Scale.
Arash Ahmadian*, Saurabh Dash* Hongyu Chen*, Bharat Venkitesh, Stephen Gou, Phil Blunsom, Ahmet Üstün and Sara Hooker.
Neural Information Processing Systems (NeurIPS), 2023.
Associative Memory Augmented Asynchronous Spatiotemporal Representation Learning for Event-based Perception.
Uday Kamal*, Saurabh Dash* and Saibal Mukhopadhyay.
International Conference on Learning Representations (ICLR), 2023.
Sequence Approximation using Feedforward Spiking Neural Network for Spatiotemporal Learning: Theory and Optimization Methods.
Xueyuan She, Saurabh Dash and Saibal Mukhopadhyay.
International Conference on Learning Representations (ICLR), 2022.
Temporal Point Process Modelling using Recurrent Graph Network.
Saurabh Dash, Xueyuan She and Saibal Mukhopadhyay.
International Joint Conference on Neural Networks (IJCNN), 2022.
Unsupervised Hebbian Learning on Point Sets in Starcraft II.
Beomseok Kang, Harshit Kumar , Saurabh Dash and Saibal Mukhopadhyay.
International Joint Conference on Neural Networks (IJCNN), 2022.
A Heterogeneous Spiking Neural Network for Unsupervised Learning of Spatiotemporal Patterns.
Xueyuan She, Saurabh Dash and Saibal Mukhopadhyay.
Frontiers in Neuroscience, 2021.
Robust Processing-In-Memory with Multi-bit ReRAM using Hessian-driven Mixed-Precision Computation.
Saurabh Dash, Yandong Luo, Anni Lu, Shimeng Yu and Saibal Mukhopadhyay.
IEEE Transactions on Computer Aided Design (TCAD), 2021.
Physics-Incorporated Convolutional Recurrent Neural Networks for Source Identification and Forecasting of Dynamical Systems.
Priyabrata Saha, Saurabh Dash and Saibal Mukhopadhyay.
Neural Networks, 2021.
Hessian-Driven Unequal Protection of DNN Parameters for Robust Inference.
Saurabh Dash and Saibal Mukhopadhyay.
International Conference on Computer Aided Design (ICCAD), 2020.

Other Publications:

Brain-Inspired Spatiotemporal Processing Algorithms for Efficient Event-Based Perception.
Biswadeep Chakraborty, Uday Kamal, Xueyuan She, Saurabh Dash and Saibal Mukhopadhyay.
Design, Automation & Test in Europe (DATE), 2023.
Reliable Edge Intelligence in Unreliable Environment.
Minah Lee, Xueyuan She, Biswadeep Chakraborty, Saurabh Dash, Burhan Mudassar and Saibal Mukhopadhyay.
Design, Automation & Test in Europe (DATE), 2021.
Impact of HKMG and FDSOI FeFET drain current variation in processing-in-memory architectures.
Nathan Miller, Zheng Wang, Saurabh Dash, Asif Khan and Saibal Mukhopadhyay.
Journal of Materials Research, 2021.
Characterization of Drain Current Variations in FeFETs for PIM-based DNN Accelerators.
Nathan Miller, Zheng Wang, Saurabh Dash, Asif Khan and Saibal Mukhopadhyay.
IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS), 2021.
A Flexible Precision Multi-Format In-Memory Vector Matrix Multiplication Engine in 65nm CMOS with RF Machine Learning Support.
Mandovi Mukherjee, Yun Long, Jongseok Woo, Daehyun Kim, Nael Mizanur Rahman, Saurabh Dash and Saibal Mukhopadhyay.
IEEE Solid-State Circuits Letters, 2020.
Low Power Implantable Spike Sorting Scheme based on Neuromorphic Classifier with Supervised Training Engine.
Rakshit Pathak, Saurabh Dash and Mrigank Sharad.
IEEE Computer Society Annual Symposium on VLSI (ISVLSI), 2017.
Low Cost Autonomous Navigation and Control of a Mechanically Balanced Bicycle.
A. Pandey, S.Mahajan, A. Kosta, D. Yadav, V. Pandey, S. Sahay, S. Jha, S. Agarwal, A. Bhise, R. Kumar, A.Bhushan, V. Parikh, A. Lohani, S. Dash, H. Choudhary, R. Kumar, A. Sharma, A. Mondal, C.K.Sai, P.N. Vamshi
Transportation Electrification Conference (ITEC), 2015.