Matthew E. Peters

Timeline

1. April 2017

   Context embeddings for sequence tagging

   In April 2017, he is first author of a paper with Waleed Ammar and two coauthors that adds pre-trained context embeddings from bidirectional language models to sequence labeling systems. The approach reaches state-of-the-art results on named entity recognition and chunking datasets.

2. February 2018

   Deep contextualized word representations

   In February 2018, he posts Deep contextualized word representations to arXiv with Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee and Luke Zettlemoyer. The paper introduces ELMo, word vectors learned from the internal states of a deep bidirectional language model, and improves the state of the art across six NLP problems. It is published at NAACL in New Orleans in June 2018.

3. March 2018

   The AllenNLP platform

   In March 2018, he is one of nine authors of AllenNLP, a PyTorch-based platform for research on deep learning methods in natural language understanding. The open-source project is maintained by engineers and researchers at the Allen Institute for Artificial Intelligence.

4. 2018

   A NAACL best paper

   In 2018, the ELMo paper, with Luke Zettlemoyer among its seven authors, wins a best paper award at NAACL.

5. September 2019

   Knowledge bases inside BERT

   In September 2019, he is first author of Knowledge Enhanced Contextual Word Representations, a method that embeds knowledge bases into large models. After integrating WordNet and a subset of Wikipedia into BERT, the resulting KnowBert improves perplexity and downstream results on relationship extraction, entity typing and word sense disambiguation.

6. April 2020

   Longformer for long documents

   In April 2020, he coauthors Longformer: The Long-Document Transformer with Iz Beltagy and Arman Cohan. Its attention mechanism scales linearly with sequence length, which makes it easy to process documents of thousands of tokens or longer.

7. December 2023

   Co-founding Spiffy

   In December 2023, Spiffy, the Seattle startup he co-founded, comes out of stealth with a $6 million seed round from investors including Point72 Ventures and the AI2 Incubator. His background includes a Senior Research Scientist post at the Allen Institute, focused on LLMs and general purpose NLP, and earlier roles as Director of Data Science at Moz and financial engineer at Fannie Mae.

8. As of October 2026

   Chief Architect at Envive

   Spiffy, founded in 2023, now operates as Envive, and he is its co-founder and Chief Architect, heading the effort to scale its machine learning and LLM technologies. Envive announces a $15 million Series A round led by Fuse VC, bringing total funding to $20 million.