Adrian Celaya

Software Engineer @ Google

I work on developing machine learning and AI solutions to advance consumer health and well-being. I earned my Ph.D. in Computational and Applied Mathematics from Rice University, combining machine learning with traditional applied mathematics to address challenges in medical imaging, consumer health, and geophysics.

Previously, I served as the Information System Security Manager aboard the USS Carl Vinson, leading the ship's cybersecurity program.

Download CV

Research Interests

AI for Health Medical Imaging Segmentation Scientific Machine Learning Health Foundation Model Training

Software

MIST Open in Colab

Medical Imaging Segmentation Toolkit (MIST)

A simple, scalable, end-to-end framework for 3D medical image segmentation — from raw NIfTI files to trained models and evaluated predictions, with sensible defaults out of the box.

Flexible, easy alternative to nnU-Net
CPU, CUDA & ROCm support
Extensible loss & architecture registry
Try it in Colab — no setup required

Python PyTorch Open Source

GitHub Docs PyPI Docker Paper

MISFIT

Medical Imaging Semantic Foundation Toolkit (MISFIT)

A simple, scalable, end-to-end framework for pretraining 3D medical imaging foundation models with masked autoencoders. Give it unlabeled NIfTI files and it produces a pretrained encoder — no labels required.

No labels required
3D-native SwinUNETR encoder
Single-GPU to multi-node scaling
Encoder is compatible with MIST

Python PyTorch Open Source

GitHub Docs PyPI Docker

Selected Publications

View Google Scholar →

MIST: A Simple and Scalable End-To-End 3D Medical Imaging Segmentation Framework

A. Celaya, et. al.

Accepted BraTS 2024 Challenge @ MICCAI 2024

Paper

Pre- and Post-Treatment Glioma Segmentation with the Medical Imaging Segmentation Toolkit

A. Celaya, T. Netherton, D. Schellingerhout, C. Chung, B. Riviere, D. Fuentes

International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2025)

Paper

A Generalized Surface Loss for Reducing the Hausdorff Distance in Medical Imaging Segmentation

A. Celaya, B. Riviere, D. Fuentes

arXiv preprint (2023)

Paper

Training Robust T1-Weighted Magnetic Resonance Imaging Liver Segmentation Models...

A. Celaya, et. al.

Scientific Reports (2024)

Paper

Adaptive Collocation Point Strategies For Physics Informed Neural Networks via the QR Discrete Empirical Interpolation Method

A. Celaya, D. Fuentes, B. Riviere

arXiv preprint (2025)

Paper

Solutions to Elliptic and Parabolic Problems via Finite Difference Based Unsupervised Small Linear CNNs

A. Celaya, K. Kirk, D. Fuentes, B. Riviere

Computers & Mathematics with Applications (2024)

Paper

PocketNet: A Smaller Neural Network For Medical Image Analysis

A. Celaya, et. al.

IEEE Transactions on Medical Imaging (2022)

Paper

Get In Touch

San Francisco, CA

aecelaya@google.com