Brian Staber

I work as a machine learning engineer at Safran on applied AI systems for aerospace and defense applications. My background combines computational physics, applied statistics, and machine learning, with a focus on building practical solutions for physics-based and industrial problems.

Over the past eight years, I have worked on surrogate modeling and data-driven approaches operating on complex engineering data such as meshes, fields, and graphs, as well as on computer vision methods for industrial anomaly detection. My work involves developing end-to-end ML pipelines, from dataset preparation and model training to evaluation and deployment, often in distributed or HPC environments.

I mainly work with deep learning models (graph neural networks, neural operators, transformers), as well as kernel and Gaussian process methods, and I pay particular attention to uncertainty and model reliability in real-world settings. Some of this work has resulted in peer-reviewed publications at top-tier machine learning conferences, including NeurIPS and AISTATS.

I enjoy learning new tools and approaches, collaborating with domain experts, and contributing to robust scientific software. Most of my development work is in Python and C++, and involves standard software engineering practices such as testing, continuous integration, and reproducible workflows.