Andreas Steiner
Research Engineer at Google
Zurich, Zurich, Switzerland
Summary
🤩
RockstarAndreas Steiner is a research engineer at Google Brain in Zürich with a decade of hands-on experience building and improving ML training pipelines and model code. He contributes to prominent open-source projects like Flax and Vision Transformer, focusing on practical engineering—TPU/GPU compatibility, Optax optimization, metrics logging, and extended model configurations for ViT and ResNet-ViT variants. His work on TensorFlow workshops shows a strong bent for data preprocessing and making research-ready notebooks Colab-friendly for workshops and teaching. Andreas blends research-grade model development with pragmatic engineering tradeoffs that make experiments reproducible and runnable on common cloud and TPU setups. Colleagues would note his knack for smoothing the bridge between research prototypes and usable examples that accelerate adoption by the community.
10 years of coding experience
Github Skills (19)
python10
optim10
machine-learning10
flax10
data-preprocessing10
deep-learning10
trainings10
tensorflow10
tensorflow-datasets10
computer-vision10
jupyter-notebook10
jax10
opt10
modeling10
machine-learning-models9
Programming languages (4)
Github contributions (5)
google-research/vision_transformer
Oct 2020 - Jan 2023
Role in this project:
ML Engineer
ML EngineerContributions:29 reviews, 148 commits, 66 PRs in 2 years 3 months
Contributions summary:Andreas forked the vision_transformer repository and added a significant number of features related to model configuration and training. They added configurations for various ViT model sizes (L/16, H/14), as well as a ResNet-based ViT model. Additionally, they introduced flags for shuffling data, setting the dataset directory, and specifying the optimization data type (e.g., bfloat16). These changes indicate involvement in model development and training process.
tensorflow/workshops
Jan 2018 - Feb 2020
A few exercises for use at events.
Role in this project:ML Engineer
ML EngineerContributions:22 commits, 21 PRs, 10 pushes in 2 years 1 month
Contributions summary:Andreas contributed to the AMLD (Advanced Machine Learning Days) workshop within the tensorflow/workshops repository. Their work involved adding and modifying notebooks related to the Quickdraw dataset, including data loading, data inspection, rasterization, and data preparation for machine learning models. They also updated the notebooks for compatibility with Colab and TensorFlow 2. The user's changes demonstrate a focus on data preprocessing and preparation for machine learning tasks.
javaevents
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.