Benjamin Englard

Education

Thiel Fellowship: June 2014 - Present

University of Michigan: Computer Science Engineering & Mathematics

MyTE is building VR/AI based training tools. After beginning as virtual classroom AI for teacher training, MyTE is now expanding into a platform for building novel training applications for any industry.

For MyTE I have built and deployed novel and potentially patentable dialogue systems, browser based graphics, vision, and simulation systems, and content management systems.

Research

My main research interest at this time is: novel uses of/architectures for neural networks, with applications to natural language processing and computer vision. I am specifically interested in trainable memory models, recurrent models of vision, and deep reinforcement models.

Relevant work:

MyTE_NLP - The NLP library built on torch7 used at MyTE. It includes sublibraries for building recurrent neural networks models, reinforcement learning models, and a general use nlp toolkit.
symtorch - a graph computation library, capable of easy expression of neural nets.
symnn - a neural network library built on top of symtorch.
Inveling - I trained a LSTM-convnet (encoder-decoder) neural network architecture to map text input into image output. In the process, I built a neural network library on top of theano.
ccnn - I (attempted) to train a recurrent neural network to be a C compiler.
vidAIo - set of natural language processing and computer vision technologies that allow for search within both the audio and frame content of a video.
ConvNetPy - neural nets in Python
RustyNet - neural nets in Rust

Florida International University High Performance Database Research Center: Designed and implemented a Spatial Autocorrelation system, the first prototype of what will eventually become the TerraFly Spatial Analysis toolset.

University of Florida Advanced Computing and Information Systems Laboratory: Designed, implemented, unittested, and simulated the message recovery system for a Peer-to-Peer alternative to Twitter, Litter. Worked remotely updating the Litter web interface.

A Rhetorical Analysis Approach to Natural Language Processing: Applied a method of literary analysis to problems in Natural Language Processing and developed novel solutions to several open problems.