Multimodal foundation
RADEO.
A compact foundation for a wide world of medical images.
Trained across a diverse medical imaging corpus, RADEO learns representations that can be adapted to specialized tasks and datasets.
Medical foundation models
Foundational
intelligence for
better medicine.
Intelligence built from medical images.
Adapted to the decisions that matter.
From signal to understanding
A vision for multimodal intelligenceAcross modalities.
Starting with the heart.
Deep research.
A practical foundation.
160M
Images in RADEO’s training corpus
14
Medical imaging modalities
86M
Parameters in the RADEO backbone
01 / Our vision
Medical intelligence.
Within reach.
MaxMAI’s vision is to make advanced medical intelligence accessible across institutions and care settings.
We are building AutoMAX, an efficient platform that combines medical foundation models and automated, task-specific optimization to help hospitals, health systems, research institutions, and developers build and adapt medical AI to their own data and needs—with less compute and development effort.
Our first clinical focus is cardiac intelligence, with the goal of developing FDA-cleared decision-support tools for hospitals. Through integration with point-of-care ultrasound, we aim to extend these capabilities to rural and underserved communities with limited access to specialists.
02 / Our models
Broad intelligence.
Focused expertise.
A shared foundation across medical imaging. Specialized models for the clinical questions that need a closer look.
Explore the research
Cardiac foundation
ECHO.
A deeper understanding of the heart, in motion.
Learning from multi-view echocardiography, ECHO captures both spatial features and temporal patterns to identify subtle differences that inform difficult cardiac decisions.
03 / Research in focus
The hard questions
are where
we begin.
Our research connects advances in representation learning with the distinctions that matter in medical imaging.
Request research detailsCompact model.
Strong benchmark results.
In our research evaluations, RADEO outperformed MedGemma on 14 medical imaging benchmarks, using an 86M-parameter backbone.
Medical image classificationModel efficiency
Distinguishing two
look-alike conditions.
In an echo-only research study, our model classified STEMI versus Takotsubo syndrome more accurately than participating clinicians.
STEMI vs. TakotsuboEchocardiography-based evaluation
Study-specific research findings. Methods, cohorts, and detailed results are available on request. Models are not FDA-cleared for clinical use.
04 / Our approach
Better models start
with better objectives.
Medical AI needs to work on the cases that are rare, ambiguous, and easy to miss. Our work starts with how models learn.
01
Train more efficiently.
We develop algorithms that reduce the cost of foundation-model training. Our earlier research outperformed OpenAI CLIP on evaluated benchmarks using 1/17 the compute and 1/10 the training samples processed.
02
Optimize what matters.
We develop objectives around AUROC, AUPRC, and partial AUC to address class imbalance and clinically relevant operating ranges.
03
Make adaptation practical.
Compact models and focused fine-tuning create a path toward deployment within the constraints of healthcare.
Our optimization foundationLibAUC
Building on open-source optimization research to make specialized medical AI easier to develop.
Explore LibAUC (opens in a new tab)05 / The path ahead
Start with a clear need.
Build toward a wider impact.
01Enabling platform
AutoMAX
Develop a pretraining and fine-tuning platform that combines medical foundation models with automated, task-specific optimization for efficient adaptation to local data and needs.
02First clinical focus
Cardiac intelligence
Develop and validate ECHO for defined cardiac decision-support tasks, with the goal of FDA clearance.
03Broader access
Point-of-care ultrasound
Integrate validated cardiac intelligence with point-of-care ultrasound to support rural and underserved communities with limited specialist access.
Let’s build what comes next
Better medicine
takes all of us.
For hospitals, health systems, research institutions, and developers ready to build more accessible medical AI.
Start a conversation hello@maxmai.com