Sebastian "Sebo" Diaz

I am an MIT PhD candidate working with Elfar Adalsteinsson and Polina Golland on machine learning.

My interests include generalization, synthetic data, representation learning, and numerical methods.

I am also a Tech Venture Fellow at Engine Ventures.

You can reach me at sdd@mit.edu or sebodiaz@csail.mit.edu.

Publications

Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It
Sebo Diaz, Polina Golland, Elfar Adalsteinsson, Neel Dey
NeurIPS 2026

[project] [arXiv] [code]

Summary

Invariance alone can limit domain generalization. MaskGen combines invariant and domain-specific information for sample-efficient biomedical segmentation across domains.
Robust Fetal Pose Estimation across Gestational Ages via Cross-Population Augmentation
Sebo Diaz, Benjamin Billot, Neel Dey, Molin Zhang, Esra Abaci-Turk, Ellen Grant, Polina Golland, Elfar Adalsteinsson
MICCAI 2025

[paper] [arXiv] [code]

Summary

Cross-population augmentation improves pose estimation in younger fetuses using only older fetuses for training. Our method achieves state-of-the-art results on a challenging clinical dataset.
Design of Novel RF Pulse for Fetal MRI Refocusing Trains using Rank Factorization (SLfRank) to Reduce SAR and Improve Image Acquisition Efficiency
Sebo Diaz, Yamin Arefeen, Borjan Gagoski, Ellen Grant, Elfar Adalsteinsson
ISMRM 2022

[paper]

Summary

SLfRank optimizes RF pulses to accelerate fetal MRI by 22% over the industry standard. The approach also applies to other turbo spin echo sequences.
Fetuses Made Simple: Modeling and Tracking of Fetal Shape and Pose
Yingcheng Liu, Peiqi Wang, Sebo Diaz, Benjamin Billot, Esra Abaci Turk, Ellen Grant, Polina Golland
MICCAI 2025

[paper] [arXiv] [code]

Summary

We introduce a 3D articulated fetal body model based on SMPL to track shape and pose. It captures body structure and motion beyond isolated keypoints or segmentations.
Stochastic-offset-strategy enhanced RF pulse optimization with auto-differentiation
Molin Zhang, Nicholas Arango, Sebo Diaz, Jacob White, Elfar Adalsteinsson
ISMRM 2024

[paper]

Summary

Random spatial offsets improve RF pulse optimization with automatic differentiation. This approach outperforms fixed-point representations at coarse spatial resolutions.