Udit Gupta

I am a computer systems and architecture researcher. My research interests focus on improving the performance, efficiency, and environmental sustainability of computer systems and architecture by co-designing solutions across the computing stack.

My research optimizes the at-scale deployment of deep learning based personalized recommendation in order to enable next generation recommendation systems. Personalized recommendation is central to many Internet services we use on a daily basis including social media, search, video and movie streaming, and e-commerce platforms. Our work demonstrates that recommendation not only accounts for a majority of all AI cycles in data centers but also poses unique challenges to efficient execution that limits new modeling capabilities. Tackling these challenges we outline an in-depth analysis and characterization of the architectural implications of recommendation models across production data centers. Building on the characterization, we accelerate recommendation systems with  novel near memory processing systems, inference schedulers, and data center-scale specialized hardware for a wide set of neural recommendation models. Some of these solutions have been deployed in production data centers, demonstrating performance improvements at-scale. To enable future work into datacenter scale recommendation inference, we have open-sourced the proposed infrastructure, DeepRecSys and host the Personalized Recommendation Systems and Algorithms (PeRSonAl) workshop.

Given the growing demands for AI, emerging applications, and computing overall, I am also passionate about investigated new techniques to enable environmentally sustainable computing. We show that sustainable mobile and data-center scale computing requires reducing both operational energy consumption and embodied carbon from hardware manufacturing (Chasing Carbon). From a computing perspective this means we must rethink research and development across the computing stack (applications, systems, hardware, and devices) in order to build sustainable computer systems of the future. To this end, we host the Computing Landscapes for Environmental Accountability and Responsibility (CLEAR) workshop at ISCA 2021.

My past work has also included designing hardware accelerators for sparse RNNs, building and integrating a DNN accelerator into mobile SoC taped-out in 16nm, lossy compression techniques for DNNs, and quantifying the fault tolerance of DNNs.

I am an active steering committee member of the Computer Architecture Student Association, where I hope to build a more diverse and inclusive student community in computer architecture.

My work has received an IEEE MICRO Top Picks Honorable Mention (2020), as well as best paper nominations at the Parallel Architectures and Compilation Techniques (PACT 2019) and Design Automation Conference (DAC 2018).

I am a PhD student studying Computer Science in the Harvard Architecture, Circuits and Compilers group at Harvard University working with Professor David Brooks and Professor Gu-Yeon Wei. I have also been fortunate to collaborate with Facebook AI Research's (FAIR) SysML team where I am advised by Dr. Carole-Jean Wu.

Harvard University

PhD in Computer Science (2016-Present)
Masters of Science in Computer Science (2020)

Cornell University

Bachelor of Science, GPA 4.00 (2012-2016)
Major: Electrical and Computer Engineering
Minor: Computer Science

For a complete and up to date list of publications please visit Google Scholar.

RecPipe: Co-designing Models and Hardware to Jointly Optimize Recommendation Quality and Performance
Udit Gupta, Samuel Hsia, Jeff Zhang, Mark Wilkening, Javin Pombra, Hsien-Hsin S. Lee, Gu-Yeon Wei, Carole-Jean Wu, David Brooks
To appear in IEEE/ACM International Symposium on Microarchitecture (MICRO 2021)
[ArXiv] [Github] [MICRO 2021 Talk] [MICRO 2021 Lightning Talk] [PDF]
[Artifact] (available, functional, and reproducible)

RecSSD: Near Data Processing for Solid State Drive Based Recommendation Inference
Mark Wilkening, Udit Gupta, Samuel Hsia, Caroline Trippel, Carole-Jean Wu, Gu-Yeon Wei
International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS 2021)
[ArXiv]

Chasing Carbon: The Elusive Environmental Footprint of Computing
Udit Gupta, Young Geun Kim, Sylvia Lee, Jordan Tse, Hsien-Hsin S. Lee, Gu-Yeon Wei, David Brooks, Carole-Jean Wu
IEEE International Symposium on High-Performance Computer Architecture (HPCA 2021)
Featured by: Harvard Gazette, Tech @ Facebook, Bloomberg Green
[PDF] [Slides] [HPCA 2021 Long Talk] [HPCA 2021 Short Talk]

DeepRecSys: A System for Optimizing End-To-End At-scale Neural Recommendation Inference
Udit Gupta, Samuel Hsia, Vikram Saraph, Xiaodong Wang, Brandon Reagen, Gu-Yeon Wei, Hsien-Hsin S. Lee, David Brooks, Carole-Jean Wu
International Symposium on Computer Architecture (ISCA 2020)
[PDF] [Slides] [ISCA 2020 Talk] [Github] [PeRSonAl Tutorial Talk]

RecNMP: Accelerating Personalized Recommendation with Near-Memory Processing
Liu Ke, Udit Gupta, Carole-Jean Wu, Benjamin Youngjae Cho, Mark Hempstead, Brandon Reagen, Xuan Zhang, David Brooks, Vikas Chandra, Utku Diril, Amin Firoozshahian, Kim Hazelwood, Bill Jia, Hsien-Hsin S. Lee, Meng Li, Bert Maher, Dheevatsa Mudigere, Maxim Naumov, Martin Schatz, Mikhail Smelyanskiy, Xiaodong Wang
International Symposium on Computer Architecture (ISCA 2020)
[PDF]

Architectural Implications of Facebook’s DNN-based Personalized Recommendation
Udit Gupta, Carole-Jean Wu, Xiaodong Wang, Maxim Naumov, Brandon Reagen, David Brooks, Bradford Cottel, Kim Hazelwood, Bill Jia, Hsien-Hsin S. Lee, Andrey Malevich, Dheevatsa Mudigere, Mikhail Smelyanskiy, Liang Xiong, Xuan Zhang
IEEE International Symposium on High-Performance Computer Architecture (HPCA 2020)
Selected as Honorable Mention for IEEE MICRO's Top Picks
Featured by: Facebook Research
[PDF] [Slides] [Github] [PeRSonAl Tutorial Talk]

MASR: A Modular Accelerator for Sparse RNNs
Udit Gupta, Brandon Reagen, Lillian Pentecost, Marco Donato, Thierry Tambe, Alexander Rush, Gu-Yeon Wei, David Brooks
Parallel Architectures and Compilation Techniques (PACT 2019)
Best Paper Nominee
[PDF], [Slides], [ArXiv]

Facebook, Inc.

Facebook AI Research (FAIR) Visiting Research Scientist	January 2021 - Present
Facebook AI Research (FAIR) Intern	January 2020 - December 2020
AI Infrastructure Research Intern	September 2018 - January 2020

Algo Logic Systems Inc.

Hardware Design and Verification Engineer

May 2015 - August 2015

Harvard CS 290	Fall 2020
Harvard CS 141	Spring 2019
EdX MOOC: The Computing Inside Your Smart Phone	Summer 2014
Cornell ECE 2300: Introduction to Digital Logic and Computer Organization	Spring & Fall 2014, Spring 2015
Cornell CS 3420 / ECE 3140: Embedded Systems	Spring 2016

Middle-school and high-school outreach

Rainstorm, Sustainable Computing	Spring & Summer 2021
Splash, Computers Don't Byte!	Fall 2014 and Spring 2015

CLEAR at ISCA 2021 Co-Founder and Co-Organizer	2021
NOPE at ASPLOS 2021 Co-Organizer	2021
JOURNE at MLSys 2021 Co-Founder and Co-Organizer	2021
PeRSonAl at MLSys 2021 Co-Founder and Co-Organizer	2021
PeRSonAl at ISCA 2020 Co-Founder and Co-Organizer	2020
PeRSonAl at ASPLOS 2020 Co-Founder and Co-Organizer	2020
NOPE at ASPLOS 2019 Co-Organizer	2019

Computer Architecture Student Association (steering committee member)	2020 - Present
3C (Cultural Competence in Computing) Fellow	2021 - 2022
Harvard SEAS Graduate Council Member	2020 - Present
Harvard SITN Lecture Director	2018 - 2019
Harvard SITN Blog Editor	2018 - 2019
Cornell IEEE President	2015 - 2016
Cornell IEEE Corporate Director	2013 - 2015