Shraman Ray Chaudhuri
shramadasflkjhan [at] mdfhsdjafit [dot] 1432@#$2 edu

I'm currently an M.Eng. student at MIT, where I work with Professor Josh Tenenbaum on computer vision.

Previously, I received my B.S. in computer science with a minor in math from MIT, where I worked with Professor Nir Shavit on computational connectomics.

Outside MIT, I've worked on some fun research projects at D.E. Shaw Research on molecular dynamics (MD) simulation, and SpaceX on machine learning for large-scale telemetry analysis.

Resume  |  CV  |  LinkedIn  |  Code

Deep Tensor Convolution on Multicores (paper)
with David Budden, Alex Matveev, Shibani Santurkar, Nir Shavit
International Conference on Machine Learning (ICML), 2017

We derive a suite of Winograd-style Fast Fourier Transforms to minimize computation for high-dimensional convolutions. Our algorithm runs faster than other deep learning libraries (e.g. Caffe, TensorFlow) for both inference and training on CPU.


More to be uploaded soon!


Auto-encoding Variational Bayes with Extensions (paper) (code)
6.882: Advanced Bayesian Modeling and Inference, 2016

Explored variational autoencoders for image compression. Derived and implemented Natural Gradients to optimize the variational objective. Measured the effect of various loss functions and parameterizations of the latent space on reconstruction quality.


Solving Traffic Flow Problems with Godunov's Flux Method (paper) (code)
6.339: Numerical Methods for Partial Differential Equations, 2016

Derived and implemented a PDE solver for multi-lane traffic models using high-resolution finite volume methods.


Real-time Activity Recognition with Google Glass (paper)
6.869: Advances in Computer Vision, 2015

Developed a fast algorithm for 3-way activity recognition from low-resolution, 30fps video.


[TA] 6.046 - Design & Analysis of Algorithms - Fall 2016

[TA] 6.046 - Design & Analysis of Algorithms - Spring 2017

[Head TA] 6.046 - Design & Analysis of Algorithms - Fall 2017


[TA] 6.S191 - Introduction to Deep Learning - Winter 2017

Website creds to Jon Barron