Matthew Peters
Co-Founder at Allen Institute for Artificial Intelligence (AI2)
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolMatthew Peters is a co-founder and chief architect building Outcome-Oriented Models at Envive/Spiffy AI, focused on AI systems that continuously improve toward user-centric goals. With 11+ years bridging research and product, he came from AI2 where he led foundational NLP work (ELMo, KnowBert, Longformer) and helped found AllenNLP. He combines deep academic training—a PhD in Applied Mathematics—with hands-on ML engineering, contributing production-ready model implementations and tooling in major open-source projects like allenai/bilm-tf and allennlp. Prior roles as Director of Data Science at Moz and senior quantitative positions in finance show a track record of putting large-scale, high-throughput models into production and influencing product strategy. He has a practical knack for release engineering and deployment (e.g., packaging and CI fixes on dragnet) that complements his research pedigree. Based in Seattle, he blends curiosity-driven research with pragmatic engineering to ship performant AI systems that improve measurable outcomes.
11 years of coding experience
13 years of employment as a software developer
Ph.D. Applied Mathematics, Ph.D. Applied Mathematics at University of WashingtonB.S. Mathematics, B.S. Mathematics at Penn State University
Github Skills (18)
pytorch10
python10
machine-learning10
cicd10
deep-learning10
tensorflow10
natural-language-processing10
nlp10
devops9
data-processing9
pylint8
bash8
hdf8
git7
mypy7
Programming languages (3)
Github contributions (5)
allenai/bilm-tf
Sep 2017 - Oct 2018
Tensorflow implementation of contextualized word representations from bi-directional language models
Role in this project:
ML Engineer
ML EngineerContributions:70 commits, 20 PRs, 17 pushes in 1 year
Contributions summary:Matthew's contributions focused on implementing and improving the core functionality of a bidirectional language model (biLM) in TensorFlow. They added a batcher class to handle data processing and designed the architecture, including character-level and token-level embedding implementations. Furthermore, the user integrated an ELMo (Embeddings from Language Models) component into the framework and built scripts to test, train, and dump weights, indicating involvement in the model's lifecycle.
nlprepresentationsword-embeddingsword-representationsbi-directional
dragnet-org/dragnet
Dec 2016 - Nov 2022
Just the facts -- web page content extraction
Role in this project:DevOps Engineer
DevOps EngineerContributions:2 releases, 39 commits, 39 PRs in 5 years 11 months
Contributions summary:Matthew focused on improving the project's build and deployment processes. They fixed the Travis CI build, modified setup.py to incorporate necessary dependencies and package configurations, and prepared the project for a v2 release. Their contributions included using setuptools instead of distutils and bumping the PyPI version, indicating involvement in package management and release procedures.
extractionpythoncontent-extractionfactsweb-page
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.