Ellery Buntel's Personal Site/ Project Portfolio

Ellery Buntel's Homepage

Welcome!

About me

Hello! I’m Ellery Buntel, a computer scientist specializing in computational neuroscience and brain comptuer interfaces.

I am currently on a period of leave from a computer science PhD program at Tufts University, but my work in brain computer interfaces continues.

My work today is focused on my startup, Telepath, which is creating non-invasive brain computer interfaces for communication. Our goal is to understand the language of the brain in unprecedented detail by helping solve speech paralysis for people with severe communication disorders. Please check out our website, Telepath, for more details.

I was recently a member of the Augmentation Lab’s 2023 augmentation residency, where we developed fundamental technologies in the field of augmentation. Our projects included a collaboration with MIT’s Media Lab to build a system for reconstructing images from solely human neural data.

My research is focused on brain computer interfaces and machine learning systems for interpreting neural data. In practice, this means I do a lot of deep learning work, but I also have experience with other emerging technologies like reinforcement learning and cloud computing.

This website is meant to showcase a few of my major projects as well as the technologies used. Check below for some quick overviews of the projects, and look at each project page for additional details.

If you have any questions about the me / my projects, feel free to get in touch via the Contact page.

Projects

RL - Using fNIRS neural data to Improve RL Agent Learning

A project experimenting with the use of fNIRS (functional near infrared spectroscopy, a brain imaging technique) as a physiological input to a reinforcement learning agent that learns from human demonstrations.

CNN - Using Bayesian Deep Learning to Improve CNN Image Recognition Accuracy

This project used bayesian deep learning techniques in conjunction with a standard image recognition convolutional neural network (CNN) to produce a network architecture that has superior classification accuracy without requiring an increase in training data.

Timeseries - Some of my Major Projects Involving Timseries Data

A few projects related to timeseries data that I have done as part of my brain computer interfaces research. One is a web application for exploring similar subsections of a timeseries dataset, and the other is an application for running machine learning jobs on neural data.