Myle Ott
Founding-team member at Thinking Machines Lab; first author of fairseq and co-author of OPT and RoBERTa
- Founding team, Thinking Machines Lab12
Myle Ott is an AI researcher on the founding team at Thinking Machines, and earlier was part of the founding team at Character AI. Before that, Ott led large language model efforts at Meta AI (FAIR). Ott is first author of the 2019 paper on fairseq, an open-source sequence modeling toolkit from Facebook AI Research, and a co-author of RoBERTa and of OPT, Meta AI's 2022 paper on Open Pre-trained Transformer language models. Ott's homepage also lists FSDP (Fully Sharded Data Parallel) and Tinker, Thinking Machines' API for fine-tuning language models, among selected work.12356
Work
fairseq (external site: github.com)Open-source code
First author of the 2019 fairseq paper, listed with Facebook AI Research, describing the open-source PyTorch-based toolkit for training translation, summarization, and language models; the repository is released under the MIT license.341
OPT: Open Pre-trained Transformer Language Models (external site: arxiv.org)Published research
Co-author of Meta AI's 2022 OPT paper; Ott's homepage lists it among selected work.51
RoBERTa: A Robustly Optimized BERT Pretraining Approach (external site: arxiv.org)Published research
Other documented roles
Roles the sources document but that are not confirmed as current.
- Founding team, Character AI1
- Led large language model efforts, Meta AI (FAIR)135
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.