Matt Kelley
Assistant Professor at George Mason University
Virginia, United States
Summary
👤
Senior🎓
Top SchoolMatt Kelley is an Assistant Professor of Linguistics and computational phonetician with eight years of experience bridging phonetics research, teaching, and machine learning. Based at George Mason University after postdoctoral work and lab leadership at the University of Washington, he studies speech with a strong computational toolkit informed by a PhD from the University of Alberta. He brings practical ML engineering experience to linguistic problems, including contributing optimized CPU and GPU implementations of CTC loss to the popular Flux.jl ecosystem. An experienced instructor, he has repeatedly served as principal instructor for undergraduate phonetics and linguistics courses and directed phonetics labs. His background in Spanish, English (linguistics), and a CS minor give him a multidisciplinary perspective that shapes reproducible, open-source approaches to speech science.
8 years of coding experience
5 years of employment as a software developer
Doctor of Philosophy - PhD, Linguistics, Doctor of Philosophy - PhD, Linguistics at University of AlbertaBachelor of Arts (B.A.), Spanish, English (linguistics emphasis), minor in computer science, Bachelor of Arts (B.A.), Spanish, English (linguistics emphasis), minor in computer science at Boise State University
English, Spanish, Portuguese
Github Skills (8)
neural-network10
machine-learning10
deep-learning10
loss10
data-science10
ctc10
julia9
cuda8
Programming languages (4)
Github contributions (5)
FluxML/Flux.jl
Jun 2018 - Jan 2021
Relax! Flux is the ML library that doesn't make you tensor
Role in this project:
Data Scientist
Data ScientistContributions:9 reviews, 40 commits, 4 PRs in 2 years 7 months
Contributions summary:Matt contributed to the Flux.jl library by adding and modifying functions related to the connectionist temporal classification (CTC) loss, which is crucial for sequence-to-sequence tasks. Their work involved implementing a GPU implementation of CTC loss, optimizing the CPU version, and integrating it into the Losses module. This included refactoring of existing code, correcting syntax, and adding test cases.
ml-librarythe-human-braindata-sciencedeep-learningneural-networks
maetshju/MFCC.jl
May 2018 - Oct 2022
Mel Frequency Cepstral Coefficients calculation for Julia
Contributions:18 pushes, 8 branches in 4 years 6 months
software-defined-radiogradientsphoneticsgmmusrp
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.