Featured project01
SarataniAI
AI-assisted cervical cancer screening, built for earlier detection and more accessible clinical decision support.
Computer visionCervical screening
Visit SarataniAI Physician · Researcher · Builder
Dr. Sang'udi
Sang'udi
Medical Doctor · AI Researcher · Health-Tech Entrepreneur
Building intelligent technologies for earlier diagnosis, better clinical decisions, and more accessible healthcare.
Bristol, United Kingdom
Now
Building clinically useful AI for resource-constrained healthcare environments, while completing an MSc in Artificial Intelligence for Medicine and Health at the University of Bristol.
Updated 202601About
A physician who builds
for the real world.
I work at the intersection of medicine, machine learning, medical imaging, and global health. My interest in AI grew from seeing how delayed diagnosis, limited specialist capacity, and unequal access to healthcare technology shape patient outcomes.
Today, I develop practical systems that can support healthcare professionals in settings where infrastructure, connectivity, and compute cannot be assumed. The goal is not technology for its own sake, but tools that are useful, explainable, and close enough to clinical reality to matter.
Read my journeyCurrently exploring
Multimodal medical AIVision-language modelsFoundation models for medicineExplainable AIImplementation in African health systems
02Featured project
SarataniAI
Building AI for
earlier detection.
Cervical cancer is highly preventable when detected early, but screening systems in many low-resource environments face limited pathology capacity, long diagnostic turnaround times, and unequal access to specialists.
SarataniAI applies computer vision and deep learning to cervical cytology and visual screening images, supporting healthcare professionals with faster screening and diagnostic decision support.
Explore SarataniAI
Computer vision · Deep learning · Clinical decision support
Pap smear classificationBethesda SystemVIA image analysisExplainable AI + Grad-CAM
03Research
From promising algorithms
to useful tools.
My research explores how artificial intelligence can move from promising algorithms to clinically useful tools in real healthcare environments.
01AI for cancer diagnosisComputer vision and deep learning systems for screening and diagnostic support.02Medical imagingDeep neural networks for cytology, cervical imaging, microscopy, and clinical datasets.03Clinical AITools designed to complement clinicians and strengthen decisions, not replace judgement.04Global health AITechnologies appropriate for health systems with limited specialists, infrastructure, and compute.05Responsible AIInterpretability, validation, fairness, safety, and responsible deployment in practice.
04Selected work
Projects with a
clinical centre.
Start a conversation Digital health02
MamaCare Labour Monitor
Digital labour monitoring and clinical decision support aligned with maternal-care workflows.
Clinical AIMaternal health
View project Dataset in collection03
TANZACYT
A Tanzanian cervical cytology image dataset supporting locally grounded research in medical imaging.
DatasetMedical imaging
View project 05Journey
Medicine is the
starting point.
My work moves between clinical questions, technical systems, and the people who need them to work. Each chapter informs the next.
NowMSc Artificial Intelligence for Medicine and HealthUniversity of Bristol · Spärck AI Scholarship
CurrentCo-founder & CEO, SarataniAIsarataniai.com
TeachingAdjunct Teaching AssociateDepartment of Biomedical Engineering · MUHAS
CommunityFounder, medical student coding communityMore than 200 medical students at MUHAS
Personal philosophyTechnology matters most
Technology matters most
when it reaches the patient
who needs it.
Benchmark performance is only the beginning. In real clinical environments, cost, workflow, infrastructure, trust, explainability, and accessibility determine whether an innovation actually improves care.
06Blog
Notes on medicine
& machine learning.
More writing on LinkedIn Why earlier detection is an engineering problem
Delayed diagnosis is rarely about a single missing test. It is about workflows, capacity, and access — and why AI has to meet all three to matter.
Read articleBuilding medical AI for low-resource settings
Notes on designing computer-vision tools that assume limited specialists, unreliable connectivity, and modest compute — without compromising on safety.
Read articleFrom benchmark to bedside
A model that wins on a benchmark can still fail in clinic. What it actually takes for a promising algorithm to become a clinically useful tool.
Read article07Contact
Let's build
better healthcare.
I'm interested in collaborations across medical AI, clinical research, computer vision, global health, and responsible healthcare innovation.