Matthew Douglas
ML Engineer at Hugging Face
Lebanon, Indiana, United States
Summary
🤩
Rockstar🎓
Top SchoolMatthew Douglas is an ML Engineer with seven years of software engineering experience, recently joining Hugging Face after a decade driving full-stack and principal engineering work at PERQ. He focuses on NLP, machine learning, accessibility, and UX, blending production-grade systems design with careful attention to user-facing quality. At Hugging Face he contributes to the widely used transformers library, stabilizing bitsandbytes integration and improving 4-bit/8-bit quantization reliability across accelerator and optimizer changes. He also has deep MLOps experience from work on the bitsandbytes project, optimizing CUDA build matrices and CI/CD workflows to make k-bit quantization accessible and robust. Based in Lebanon, Indiana, Matthew pairs pragmatic engineering with an eye for performance and deployment nuances that often hide in low-level build and environment configuration. His background in full-stack development and open-source collaboration helps him bridge model research and production engineering effectively.
7 years of coding experience
14 years of employment as a software developer
BS Computer Science, BS Computer Science at Central Connecticut State University
Github Skills (18)
transformers10
pytorch10
python10
machine-learning10
cicd10
build-automation10
cuda10
quantization10
nlp10
llm9
cmake9
deep-learning9
docker8
githubaction-workflow8
tensorflow8
Programming languages (6)
Github contributions (5)
bitsandbytes-foundation/bitsandbytes
Jun 2023 - Apr 2025
Accessible large language models via k-bit quantization for PyTorch.
Role in this project:
MLOps Engineer
MLOps EngineerContributions:9 releases, 130 reviews, 90 PRs in 1 year 10 months
Contributions summary:Matthew's commits primarily focused on enhancing the build and deployment processes for the `bitsandbytes` library, which is related to LLMs and quantization. The user expanded the CUDA Toolkit version matrix used in the build system, integrated CUDA builds for new versions, and modified build workflows for improved compatibility and publishing. Their work involved adjusting paths for pip installed cmake, testing manylinux builds, and ensuring correct library loading based on environment variables, demonstrating a clear focus on continuous integration and deployment.
cudapytorch8-bitgpupruning
huggingface/transformers
Mar 2021 - Mar 2025
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:31 reviews, 8 PRs, 8 pushes in 4 years
Contributions summary:Matthew primarily contributed to fixing bugs and improving the stability of the bitsandbytes (bnb) integration within the transformers library, specifically focusing on 4-bit and 8-bit quantization. They addressed issues related to excessive CPU memory usage, test failures within bnb-based training tests, and incorrect dequantization processes. Their work ensures correct functionality with new versions of the accelerate library and the integration of new optimizers with bitsandbytes.
pythonbertspeech-recognitionstate-of-the-artflax
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.