William Buchwalter
Principal Applied Scientist at Microsoft
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolWilliam Buchwalter is a Principal Applied Scientist with 11 years of experience building and optimizing large-scale ML systems, now based in San Francisco. He co-founded AlpacaML and led product and infrastructure efforts for a diffusion-based design platform that scaled to 300K+ users and raised a $4.2M seed round. At Microsoft Research and Microsoft Turing he shipped high-performance language and vision models, improving training speed by ~10x and deploying LLMs with sub-50ms latencies in Office products. His open-source contributions span Kubeflow, Fairing, and GPU-aware Kubernetes tooling, reflecting deep expertise in distributed training, GPU orchestration, and production inference pipelines. He pairs research-grade results (co-author of a highly cited NeurIPS paper) with pragmatic engineering—automating containerized ML workflows that reduced operational friction for hundreds of teams. Notably, he’s applied that blend of research and production engineering to accelerate video and image generation model training on massive GPU clusters.
11 years of coding experience
10 years of employment as a software developer
Baccalauréat Scientifique, Spécialité Mathématiques, Baccalauréat Scientifique, Spécialité Mathématiques at Lycée Pierre-de-FermatMaster of Science (M.Sc.), Computer Science, Master of Science (M.Sc.), Computer Science at SUPINFO
French, English, Spanish
Github Skills (35)
kubernetes10
javascript10
docker10
python10
formatting10
kubeflow10
machine-learning10
cloud-infrastructure10
text-formatting10
dcos10
redux10
dockers10
azure-aks10
azure-kubernetes-service10
go10
Programming languages (10)
Github contributions (5)
angular-redux/ng-redux
Jul 2015 - Jul 2017
Angular bindings for Redux
Role in this project:
Front-end Developer
Front-end DeveloperContributions:5 releases, 189 commits, 46 PRs in 2 years
Contributions summary:William appears to have been involved in the development of the ng-redux library, an Angular binding for Redux. Their commits demonstrate the implementation of core functionalities such as connecting components to the Redux store and handling state updates. The user refactored the connector to align with an updated API, and they also added support for handling lifecycle events for better resource management. The work focused on providing a simplified interface for integrating Redux into Angular applications.
reduxangular
kubeflow/trainer
Jul 2017 - Jun 2018
Distributed ML Training and Fine-Tuning on Kubernetes
Role in this project:Backend & DevOps Engineer
Backend & DevOps EngineerContributions:27 commits, 21 PRs, 5 pushes in 10 months
Contributions summary:William primarily focused on maintaining and enhancing the Kubeflow trainer within a Kubernetes environment. Their work included fixing code indentation, running `gofmt` for code formatting, and merging branches to incorporate the latest changes. They also contributed to the integration of the TensorBoard and implemented features for parameter server setup, indicating involvement in both backend and DevOps aspects of the project.
xgboostkubernetesmachine-learningtrainingai
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.