About Me

I completed my Bachelors in Electrical Engineering and Computer Science from MIT in 2015, my M.Eng. from MIT in 2016, and am currently pursuing a PhD in Prof. Joe Paradiso's Responsive Environments group at the MIT Media Lab. My current research interests lie at the intersection of machine learning, signal processing, cognition, and audio; my thesis work entails exploring methods to build statistical models of how our brains interact with the sounds around us, and translating these models to the auditory interfaces of the future to facilitate more meaningful, compelling experiences. I was supported by the NSF Graduate Research Fellowship from 2016-2019, and am currently supported by the 2020 Apple AI/ ML Fellowship. When I'm not in the lab, I enjoy singing, playing instruments, writing, composing, and DJ-ing at MIT's college radio station.

Projects

Below are snapshots of some of the most recent projects I have worked on, both as a part of my research and otherwise.

spotify

"Cognitive Audio":
An Overview and Looking Ahead

Spring 2021

A review and opinion piece that reflects on the intersection of cognitive models and auditory interfaces, authored by myself, David Ramsay, and Clement Duhart.

Publication (preprint): Cognitive Audio Interfaces: Mediating Sonic Information With an Understanding of How We Hear -- IEEE Pervasive Computing, 2021

Active: True

spotify

Building Perceptual Metrics for Spatial Audio Applications

Spring 2020

My most recent work from an internship with Facebook Reality Labs Audio Team. More information to appear here after publication.

Publication: Under Review.

Active: True

spotify

A Perceptual Loss Metric for Limited Capacity Networks

Fall 2019

We present ideas for improving the performance of limited capacity neural networks for on-device source separation tasks using a perceptually motivated loss metric.

Publication: Using a Neural Network Codec Approximation Loss to Improve Source Separation Performance in Limited Capacity Networks -- IJCNN Special Session on Deep Neural Audio Processing, 2020

Active: True

spotify

Cognitive Audio Summarization

Spring 2019

We present early ideas on a system for automated audio summarization, as an illustration of an experience that can be constructed from cognitive models.

Publication: Under Review.

Active: True

spotify

Towards Perceptual Loss Metrics

Fall 2018

Error metrics used for training generative audio networks, typically l1 or l2 norms, do not take psychoacoustics into account. We present early ideas for designing more perceptually relevant loss metrics for deep neural networks.

Publication: Towards a Perceptual Loss: Using a Neural Network Codec Approximation as a Loss for Generative Audio Models - ACM Multimedia 2019

Active: False

stft

Modeling Auditory Memory

Fall 2017

We attempt to build statistical models to understand the fundamental causal relationships between acoustic/ gestalt features and auditory memory. David Ramsay and I have been awarded the AI Grant Fellowship for continued work on this.

Publication: HCU400 Dataset - ICASSP 2019
Publication: The Intrinsic Memorability of Everday Sounds - AES 2019

Media: NPR - Towards New Musics

Datasets: HCU400 and Memory Scores

Active: True

spotify

Machine Learning-enabled Compression

Summer 2018

How can we capitalize on the shared properties of audio to improve post-compression reconstruction, driven by machine learning? This work was done as a part of a summer internship at Spotify!

Active: False

stft

Informative Music Manipulation

Fall 2017

How can we manipulate the music we listen to everyday in a way that conveys information but remains subtle?

Publication: SoundSignaling - Ubicomp Dec. 2018

Active: False

MCC

Head in my Clouds

Fall 2020

Some (just for fun!) explorations of my listening history and sonic experience as a part of Prod. Tod Machover's "Arts at the Media Lab" class.

Demo: Arts@ML Fall 2020

Active: False

stft

Voice Hiding

Spring 2017

A simple but novel and robust technique for data hiding of critical information in cover phone audio. Work advised by Prof. Neil Gershenfeld of the Center for Bits and Atoms. Patent pending.

Active: False

MCC

VisualSoundtrack

Fall 2016 - Spring 2017

A rapid prototyping tool for soundtrack composers that approaches "style transfer" contextually with signal processing and AI tricks. Check out the demo video here.

Publication: VisualSoundtrack - ICMC 2017

Active: False

MCC

MIT Community Challenge

Fall 2016

What if we could encourage kindness and giving through gamification? We built a platform to explore this in our project from Prof. Roz Picard's Tools for Wellbeing class. Check out the video demo. Publication in CHI LWP'17

Active: False

MCC

TelTale

Fall 2016

Lend us your voice and let it be heard; share your thoughts, hopes, and fears, and let them resonate. Our voice booth installation for Prof. Tod Machover's Music in Time and Place class.

Active: False

Music & Writing

Music

Music is perhaps my biggest passion. I've been learning and singing Carnatic music since I was five or six years old, and I have always been captivated by the artform's expressive nature and lyrical beauty. It has provided me with a sense of comfort, an avenue of relaxation, and a bridge into a past rich with culture, tradition, and faith. I also enjoy singing lighter forms of classical music, and enjoy jamming with friends and family. Over the years, I've learned to play a variety of instruments including the piano, violin, harmonium, and naal (a cousin of the tabla). I also enjoy dabbling in song writing, music production, and mixing/ mastering in my spare time.

Some performances and recordings across genres: Sruthi Laya TCD 2014, MIT C-Show 2016, Someone Like You (Arr.) 2015

Writing

Many of my friends know me as perpetually having my head buried in a journal, scribbling down some thought or other. I've loved writing short stories and poetry since I was in elementary school, and after taking several fantastic creative writing classes here at MIT, I've had the opportunity to refine my writings and grow as an author. Some of my writing was awarded one of MIT's Ilona Karmel prizes, and I've written a few plays that have aired on my WMBR production, "The Daydream Company."

Assistive Technology Club & Hackathon

In my last few years here at MIT, Assistive Technologies is a space that has become particularly close to my heart. After being inspired by Professor Seth Teller and his incredible class on designing AT (6.811), I worked closely with a team of students to organize ATHack 2014, MIT's first Assistive Technologies Hackathon! And we haven't looked back since. With the completion of ATHack 2019, the hack is now in its 7th year. The aim of both the class and the hackathon is to bring together "co-designers" in the community who live with disabilities and student engineers to work towards innovative solutions.

Media: MIT News on ATHack2017, MIT News on ATHack2015, An article from the MIT Lincoln Lab, Continuing the Legacy: Assistive Technology at MIT, Perkins on ATHack2015

Interested in sponsoring, getting involved, or learning about past projects? Check out our page below.

Post-It Wall

If you could leave one piece of advice behind for the world to hear, what would it be?

Some archives of past shows are below.

Of Moments and Melodies

A whirlwind tour of the tunes and tales of the Indian sub-continent.

Some archives of past shows are below.

The Daydream Company

Celebrating the spirit of radio theater with enactments of classic plays, original scripts, and a smattering of show tunes!

Some archives of past shows will be uploaded shortly.