Joshua Meier
Co-Founder at Chai Discovery
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolJoshua Meier is a Co-Founder and machine learning leader with 14 years of experience building generative models and applied AI for biology from San Francisco. He was a platform lead for protein modeling at Facebook AI—helping ship the widely used ESM protein language models—and later led AI strategy and teams at Absci before co-founding Chai Discovery. His open-source contributions to fairseq and esm show hands-on engineering (LSTM decoders, masked LM fixes, contact-prediction APIs) and an eye for production robustness and bioinformatics needs like FASTA support. Comfortable switching between research and product, he combines a Harvard CS background and early entrepreneurial streak (CEO in high school) with a track record of translating unsupervised models into practical biotech tools.
14 years of coding experience
5 years of employment as a software developer
Academy for the Advancement of Science and Technology, Academy for the Advancement of Science and Technology at Bergen County AcademiesMaster of Science (MS) Computer Science, Master of Science (MS) Computer Science at Harvard University
English, Hebrew, Spanish
Stackoverflow
2,290reputation
218kreached
14answers
52questions
Github Skills (24)
pytorch10
artificial-intelligence10
python10
machine-learning10
sequence-modeling10
lstm10
bioinformatics10
nlp10
fairseq10
apidoc8
api8
documentation7
html6
plist6
cocoa-touch6
Programming languages (8)
Github contributions (5)
facebookresearch/fairseq
Jan 2020 - Nov 2020
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:
ML Engineer
ML EngineerContributions:26 commits, 3 comments, 1 issue in 9 months
Contributions summary:Joshua primarily contributed to the development of LSTM-based models within the fairseq framework. Their work included implementing standalone LSTM decoder language models, supporting residual connections in LSTM models, and enabling the choice of max tokens in masked LM models. Furthermore, they addressed memory leak issues in the masked LM criterion and fixed truncation in the sentence ranking task, demonstrating a focus on both model development and framework maintenance. They also added a FastaDataset, which allows the use of FASTA files which are common in bioinformatics.
pytorchnlpsequencepythontransformer-architecture
facebookresearch/esm
Sep 2020 - Dec 2020
Evolutionary Scale Modeling (esm): Pretrained language models for proteins
Role in this project:ML Engineer
ML EngineerContributions:7 commits, 10 pushes, 2 branches in 3 months
Contributions summary:Joshua primarily contributed to the development and maintenance of the ESM model, focusing on enhancements and additions. They implemented a variant prediction tutorial, showcasing the application of ESM representations in downstream tasks. Contributions included adding new features like contact prediction APIs, updating documentation, and fixing minor bugs. These changes align with the project's objective of providing pretrained language models for proteins.
scale-modelingproteinspretrained-language-modelsesmscale
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.