UROP Openings

Have a UROP opening you would like to submit?

Please fill out the form.

Submit your UROP opening

Open Source Tool for Running Large Scale Deep Learning Experiments


Term:

Spring

Department:

6: Electrical Engineering and Computer Science

Faculty Supervisor:

Pulkit Agrawal

Faculty email:

pulkitag@mit.edu

Apply by:

ASAP

Contact:

pulkitag@mit.edu

Project Description

Running large scale deep learning experiments poses interesting challenges. The goal of this project is to develop open source software that allow users to quickly iterate over experiments, visualize results etc. Once the basic framework is in place, we will explore if we can make recommendations on what experiments the user should run to debug — i.e. interactive design of experiments. Note that this project is mostly about building a system and is not “research” in the usual sense. However, it will be a very rewarding experience to learn the challenges and solutions to scaling up deep learning experiments. You have an opportunity to write software that many labs around the world may use in the future :)

Pre-requisites

- Excellent coding skills (especially python) - Expert in using git. - Basic understanding of networking. - Familiarity with docker, databases, python. - Experience in training deep learning models. - Familiarity with visualization tools is a plus. - Familiarity with cloud compute platforms like AWS, Google Cloud, Azure etc. is a plus.