Sauhaarda (Raunak) Chowdhuri

Sauhaarda (Raunak) Chowdhuri

Machine Learning Researcher and Undergraduate Student

About Me

I am a machine learning researcher, consultant, and an undergraduate student at the Massachussets Institute of Technology. I’m passionate about machine learning as a tool for effecting positive change. In my free time I enjoy reading Stoic/Buddhist philosophy, stargazing, singing Indian classical music, and walking my dog, Peanut.


  • Explainable AI
  • AI Education
  • Multi-Modal Data
  • Representation Learning
  • Cognitive Science
  • Economics


  • BSc in Electrical Engineering and Computer Science, 2023

    Massachusetts Institute of Technology



Machine Learning Researcher - Perception Team

MIT Driverless

Oct 2020 – Present Cambridge, Massachusetts
  • Designing architectures for segmentation and object detection optimized for high speed racing inference.

Assistant Researcher

MIT Computer Science and Artificial Intelligence Lab

Sep 2020 – Present Cambridge, Massachusetts
  • Developing new methods for transferable black-box adversarial perturbations for machine learning.

Machine Learning Research Consultant

MIT Lincoln Labs and MIT Environmental Sustainability Initiative

Sep 2020 – Present Cambridge, Massachusetts
  • Collaborating in a high-stakes multi-shareholder process to deliver aerial LIDAR landslide predictions to assist in the creation of environmental solutions for Colombia.

Assistant Researcher

Carnegie Mellon University - Guruswami Group

Jun 2020 – Present Pittsburgh, Pennsylvania
  • Led a team to develop efficient AI tools for music analysis, speeding up my prior independent research by nearly 800%.
  • Ongoing efforts towards first-author publication in late 2020.

Machine Learning Research Consultant

Peptide Logic

Feb 2020 – Present San Diego, California
  • Designed a pipeline for mouse behavior classification with deep learning to automate data annotation for a neuroscience drug discovery program supported by NIH Grants.
  • Ongoing efforts toward a first-author publication covering the novel techniques in this approach.

Assistant Researcher

Caltech Aerospace Robotics and Control Laboratory

Jun 2019 – Aug 2019 Pasadena, California
  • Conducted research using reinforcement learning to direct quadcopters to herd bird flocks away from airports/no fly zones, and investigated other leader-follower control problems with reinforcement learning.

Graduate Level Assistant Researcher

UC Berkeley DeepDrive

Apr 2017 – Jun 2018 Pasadena, California
  • Implemented a novel modal insertion method to allow a single deep neural network to learn several distinct “behavioral modes” of operation simultaneously for autonomous driving and published this research.
  • Adapted SqueezeNet network from ImageNet to perform regression based driving tasks.
  • Managed code repository and transferred Caffe repository to PyTorch/H5PY Training repository.
  • Migrated RC robots from TX1 to TX2 and Qualcomm Snapdragon Flight platforms using ROS control system.


Kyoto Prize Scholarship

One of three $10,000 scholarships awarded to students in San Diego and Baja California.
See certificate

Best Paper Award

This best paper award was selected by conference chairs.

Making Best Use of Data Award ($1500)

Top special award of $1500 given by GoDaddy for my novel use of data augmenetation methods at the International Science and Engineering Fair (ISEF).
See certificate

Special Award - 3rd Place ($800)

Special award of $800 given by Acoustical Society of America for my novel use of data augmenetation methods at the International Science and Engineering Fair (ISEF).
See certificate


Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization

See certificate

Convolutional Neural Networks

See certificate

Neural Networks and Deep Learning

See certificate

Structuring Machine Learning Projects

See certificate

Introduction to Computer Science and Programming Using Python (6.00.1x)

See certificate

Posts and Projects

COVID Danger Meter

Project submission for HackMIT 2020.
COVID Danger Meter

Segmenting Aerial LIDAR Data

I segment some LIDAR data from an interview with MIT’s ESI Group.
Segmenting Aerial LIDAR Data

Seeing What a Gan Cannot Generate

This is an illustrated summary of the paper Seeing What a GAN Cannot Generate
Seeing What a Gan Cannot Generate


A tour of my first python package designed to help people better organize plots for their big projects.