me at skoppula dot com  (PGP key, fingerprint / Google Scholar / GitHub)

I am currently a research engineer at Google DeepMind, working at the intersection of self-supervised visual representation learning and natural language understanding. My research interests lie across machine learning theory and applications, computer security, and computer architecture.

In 2019, I was a visiting research student at ETH Zürich on a Fulbright Research Fellowship advised by Prof. Onur Mutlu. I completed my MEng with the Energy Efficient Circuits and Systems group at MIT MTL in 2018. Part of my research was also with the Spoken Languages and Systems Group at MIT CSAIL. During my MEng, I was advised by Professor Anantha Chandrakasan and Dr. Jim Glass.

In prior years, I had the worked on NVidia's Autonomous Driving Team to build new versions of PilotNet. In 2017, I worked with the Google Acoustic Modeling research team with Prof. Khe Chai Sim. In prior summers, I worked at Yahoo and Square.

I received my B.S. in Computer Science from MIT in 2016. Andover '13. HCSSiM '10. I love painting and building racecars.

Lossless Adaptation of Pretrained Vision Models For Robotic Manipulation
Mohit Sharma, Claudio Fantacci, Yuxiang Zhou, Skanda Koppula, Jon Scholz, Nicolas Heess, Yusuf Aytar
International Conference on Learning Representations (ICLR) 2023.

Where Should I Spend My FLOPS? Efficiency Evaluations of Visual Pre-training Methods [arXiV.]
Skanda Koppula, Yazhe Li, Andrew Jaegle, Evan Shelhamer, Nikhil Parthasarathy, Relja Arandjelovic, João Carreira, Olivier J. Henaff
NeurIPS 2022 Workshop: Self-Supervised Learning, Theory and Practice (NeurIPS SSL 2022).

Hierarchical Perceiver [arXiv]
João Carreira, Skanda Koppula, Daniel Zoran, Adria Recasens, Catalin Ionescu, Olivier J. Henaff, Evan Shelhamer, Relja Arandjelovic, Matt Botvinick, Oriol Vinyals, Karen Simonyan, Andrew Zisserman, Andrew Jaegle

Object discovery and representation networks [arXiv]
Olivier J. Henaff, Skanda Koppula, Evan Shelhamer, Daniel Zoran, Andrew Jaegle, Andrew Zisserman, João Carreira, Relja Arandjelović
European Conference on Computer Vision (ECCV 2022), Oral

Perceiver IO: A General Architecture for Structured Inputs & Outputs [arXiv]
Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira
International Conference on Learning Representations (ICLR) 2022

Efficient Visual Pretraining with Contrastive Detection [pdf] [arXiv]
Olivier J. Henaff, Skanda Koppula, Jean-Baptiste Alayrac, Aaron van den Oord, Oriol Vinyals, Joao Carreira
International Conference on Computer Vision (ICCV 2021), Oral

A Deep Learning Approach for Characterizing Major Galaxy Mergers [pdf] [external]
Skanda Koppula, Victor Bapst, Marc Huertas-Company, Sam Blackwell, Agnieszka Grabska-Barwinska, et al.
NeurIPS 2020 Workshop: ML for Physical Science (NeurIPS ML4PS 2020).

Accurate, Low-Latency Visual Perception for Autonomous Racing: Challenges, Mechanisms, and Practical Solutions [pdf] [arXiv]
Kieran Strobel, Sibo Zhu, Raphael Chang, and Skanda Koppula
The IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2020).

EDEN: Enabling Energy-Efficient, High-Performance Deep Neural Network Inference Using Approximate DRAM [pdf] [arXiv]
Skanda Koppula, Lois Orosa, A. Giray Yağlıkçı, Roknoddin Azizi, Taha Shahroodi, Konstantinos Kanellopoulos, and Onur Mutlu
The 52nd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO 2019).

SMASH: Co-designing Software Compression and Hardware-Accelerated Indexing for Efficient Sparse Matrix Operations [pdf] [arXiv]
Konstantinos Kanellopoulos, Nandita Vijaykumar, Christina Giannoula, Roknoddin Azizi, Skanda Koppula, Nika Mansouri Ghiasi, Taha Shahroodi, Juan Gomez Luna, and Onur Mutlu
The 52nd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO 2019).

EcoFlow: Efficient Convolutional Dataflows for Low-Power Neural Network Accelerators [arXiv]
Lois Orosa, Skanda Koppula, Yaman Umuroglu, Konstantinos Kanellopoulos, Juan Gomez-Luna, Michaela Blott, Kees Vissers, Onur Mutlu
IEEE Transactions on Computers (Special Issue on Hardware Acceleration of Machine Learning, 2022).

Understanding Recurrent Neural State Using Memory Signatures [pdf] [arXiv]
Skanda Koppula, Khe Chai Sim, and Kean Chin
2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2018).

Energy-Efficient Speaker Identification With Low-Precision Networks [pdf] [code]
Skanda Koppula, James Glass, and Anantha P. Chandrakasan
2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2018).

Learning a CNN-based End-to-End Controller for a Formula SAE Racecar [pdf] [arXiv] [code]
Skanda Koppula and Kevin Chan
Technical Report

Applying the Residue Number System to Neural Network Inference [arXiv] [code]
Mohamed Abdelhamid and Skanda Koppula
Technical Report

Robust Predictive Bayesian Analysis for Genome-Wide Association and Expression Studies [pdf] [PubMed]
Skanda Koppula, Amin Zollanvari, Ning An, and Gil Alterovitz
In the Proceedings of the American Medical Informatics Joint Summit (AMIA 2013)

Open-Source Battery Management System for Automotive Applications [code]
MIT Formula SAE, MIT Electric Vehicle Team
The first automotive-grade battery management system with open-source firmware and PCBs.

Power-Based Side-Channel Attack for AES Key Extraction on the ATMega328 Microcontroller [pdf] [code]
Utsav Banerjee, Skanda Koppula, and Lisa Ho
Demonstrated power-based sidechannel attack on AES-ECB running on an Arduino.

A Public-Key Authentication Scheme for Controller Area Networks [pdf] [code]
Skanda Koppula, Matt Chang, and Nicolas Bravo
Design and prototyping of a scheme to secure an automotive CAN bus.

Bayesian Clustering and Topic Discovery: Adventures with Gene Expression Data [pdf] [code]
Karren Yang and Skanda Koppula
Experiments to cluster and interpret gene expression using heirarchical Bayesian models.

This site was based on this!.