Connor Leahy
US Director at ControlAI
Washington, District of Columbia, United States
Summary
🤩
Rockstar🎓
Top SchoolConnor Leahy is an AI entrepreneur and researcher with nine years of experience building and aligning large language models, currently serving as US Director at ControlAI after founding and leading Conjecture. He combines hands-on ML engineering—contributions to prominent open-source projects like EleutherAI’s GPT-Neo and a TPU-enabled GPT-2 implementation—with strategic leadership on AI safety and governance. His background includes research and engineering roles at Aleph Alpha and a Computer Science education at TUM, reflecting both applied systems work and alignment-focused research. Known in the community for bridging cutting-edge model engineering with value-aligned AI development, he also maintains an active public presence on AI policy and safety discussions.
9 years of coding experience
6 years of employment as a software developer
Computer Science, Computer Science at Technical University of Munich
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (8)
transformers10
machine-learning10
gpt10
language-model10
tensorflow10
python10
tpu10
bfd9
Programming languages (5)
Github contributions (5)
ConnorJL/GPT2
May 2019 - Nov 2019
An implementation of training for GPT2, supports TPUs
Role in this project:
ML Engineer
ML EngineerContributions:32 commits, 3 PRs, 27 pushes in 6 months
Contributions summary:Connor's commits primarily involve modifications to the `optimizers.py`, `main.py`, `predict_fns.py`, and `models/gpt2/gpt2.py` files, indicating a focus on the training and prediction aspects of a GPT2 model. They implemented custom optimizers, including AdamW and Adafactor, and integrated them with TPU support. The commits demonstrate adjustments to the model's training loop, prediction functions, and weight initialization methods, and added bfloat16 support.
deep-learningpytorchgpt2training
EleutherAI/gpt-neo
Jul 2020 - Apr 2021
An implementation of model parallel GPT-2 and GPT-3-style models using the mesh-tensorflow library.
Role in this project:ML Engineer
ML EngineerContributions:2 reviews, 51 commits, 3 PRs in 9 months
Contributions summary:Connor primarily contributed to the implementation and modification of core GPT-2 related models. Their work includes exposing local attention parameters and integrating them into the model, alongside refactoring input functions to support sampling and chunking. They also made changes to support running SELU and layer-specific attention parameters.
meshlanguage-modeltransformersmodel-parallelparallel
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.