Joe Davison
Staff AI ML Engineer at BambooHR
Lehi, Utah, United States
Summary
🤩
Rockstar🎓
Top SchoolJoe Davison is a Senior Data Scientist with a decade of machine learning and software engineering experience, currently applying his expertise at BambooHR from Lehi, Utah. He blends research-grade model development with production-focused tooling, having worked at Hugging Face and contributed to flagship open-source projects like transformers and the widely used huggingface/datasets library. His background spans academic research (Harvard S.M. in Data Science, research at the University of Utah and MIT‑IBM Watson AI Lab) and industry roles building ML systems in biotech and enterprise settings. Notably, he’s implemented core tokenization and pipeline enhancements for NLP tooling and developed genetic neural architecture search components for automated model design. Colleagues describe him as someone who moves fluidly between improving developer-facing libraries and optimizing model performance in applied domains.
11 years of coding experience
8 years of employment as a software developer
S.M., Data Science, S.M., Data Science at Harvard UniversityB.S., Computer Science, B.S., Computer Science at Brigham Young University
English, Russian
Stackoverflow
1,848reputation
502kreached
3answers
2questions
dataframe
top-5%
python
top-5%
pandas
top-5%
Github Skills (34)
tokenize10
pytorch10
caching10
genetic-algorithm10
pytest10
python10
testing10
machine-learning10
pipelining10
datasets10
keras10
tokenizer10
deep-learning10
natural-language-processing10
pipe10
Programming languages (10)
Github contributions (5)
joeddav/devol
Feb 2017 - Jul 2020
Early POC of genetic neural architecture search
Role in this project:
ML Engineer
ML EngineerContributions:65 commits, 15 PRs, 73 pushes in 3 years 4 months
Contributions summary:Joe contributed significantly to the development of a genetic neural architecture search (NAS) system within the repository. Their work focused on the implementation of core NAS functionalities, including genetic algorithm components such as population generation, selection, crossover, and mutation. The user also integrated the MNIST dataset for evaluating the performance of generated neural network architectures. Furthermore, they modified the system to optimize for the inverse of the loss function.
neural-architecture-searchdeep-learninggenetic-algorithmmachine-learningkeras
huggingface/transformers
Feb 2020 - Apr 2021
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
Role in this project:Software Engineer (Focus on Tokenization and Pipeline Enhancements)
Software Engineer (Focus on Tokenization and Pipeline Enhancements)Contributions:6 reviews, 33 commits, 48 PRs in 1 year 1 month
Contributions summary:Joe primarily contributed to improving the tokenization and pipeline functionalities within the Hugging Face Transformers library. Their work included preserving spaces in GPT-2 tokenizers, adding new methods to the PretrainedTokenizer class such as get_vocab, and ensuring consistent behavior during tokenization. They also added support for the targets argument within the fill-mask pipeline, and a zero-shot classification pipeline, reflecting a focus on enhancing core NLP tools. In addition, the user also improved test coverage.
audioinferencemachine-learning-modelsmultimodaltransformers
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.