The platform
A complete healthdata infrastructure,from paper to decision.
Muganga provides the layers a health system needs in order to own its own data at last: digitise the paper, resolve it into one record per person, protect it, and only then make something useful of it.
- layers, from the sheet of paper to the decision
- 5
- products built on the same foundation
- 3
- records being digitised at Panzi
- ≈90,000
- of the data stays the property of the institution
- 100%
Where this starts
African health data already exists. It is on paper.
Decades of care are written by hand, filed in cupboards, unreadable to any system that could do something with them. The bottleneck is not a shortage of AI models it is in the cupboard. Muganga starts at that cupboard and works forward to the clinical decision.
We are not replacing the teams already doing this work. We are building the layer that was missing underneath them.
How we go about itThe five layers
Each layer does one thing, and does all of it.
None of them can be skipped. A model reading an unstructured record gets it wrong; a structured record that has not been de-identified has no business in front of a model at all. The order is the argument.
- 01
Digitise
Physical archives are inventoried, entered and checked on site, by teams trained inside the institution. Nothing leaves the hospital to be processed somewhere else.
- 02
Structure
Four visits to four services become a single longitudinal record. That resolution not the volume is what makes an archive usable.
- 03
Govern
De-identification, the written agreement and the retention limits are settled before a model reads anything. There is no route that goes around this layer.
- 04
Reason
Models tuned to real African practice not to hospitals that look nothing like a rural health centre in South Kivu run over the structured record.
- 05
Return
The result comes back into the tools people already use, in front of someone qualified. We do not diagnose and we do not recommend treatment.
What we build on top
Three products, one foundation.
Once a health system owns its own structured data, the tools that were missing become possible. These are ours; the infrastructure underneath stays open.
Open RiposteRiposte
LiveThe epidemic response platform.
Riposte tracks an outbreak in real time, coordinates the logistics, and connects front-line providers to hospitals, health zones and the authorities.
Griot
In developmentA clinical reference assistant for African practice.
Most clinical AI is trained on hospitals that look nothing like a rural health centre in South Kivu. Griot is built the other way round.
CORETX
In developmentPharmacy stock as an early signal.
When a health zone burns through rehydration salts faster than usual, something is happening before anyone files a case report.
Why Muganga
Four reasons to trust us with this layer.
- 01
Built where it runs
Muganga is built in Bukavu, South Kivu, for institutions we know from the inside. It is not a pilot designed elsewhere and adjusted at the edges.
- 02
The whole stack, not one layer
Most offers start at the model and leave the cupboard to the institution. We take the problem by the paper and carry it all the way through.
- 03
The institution keeps ownership
Records belong to the institution holding them. All work happens under a written agreement, on de-identified data, with limits agreed in advance.
- 04
Built for the network you actually have
Power cuts and slow links are the normal condition, not an edge case. Capture continues offline and syncs when the connection comes back.
Who we build with
We started with a hospital. The platform is not only for hospitals.
Panzi is our first deployment and our proof. The same foundation serves any institution that holds or needs to read African health data, on its own terms.
- 01
Hospitals and health zones
Archive digitisation, one record per patient, and the hospital management system plugged into that record.
Archives · patient record · hospital management
- 02
Ministries and governments
National outbreak monitoring, response logistics, and reporting that runs from the health zone up to the centre.
Monitoring · logistics · reporting chain
- 03
Research and universities
De-identified longitudinal cohorts, under written agreement with the institution that holds the records. This data exists nowhere else.
Cohorts · written agreement · de-identified data
- 04
Pharma and life sciences
Real-world evidence from African practice, trial feasibility and site selection always on the institution's terms, never by buying its records.
Real-world evidence · feasibility · site selection
- 05
NGOs and humanitarian response
Response coordination, community involvement, and early signals from what the stock is actually doing.
Coordination · communities · stock signals
We do not sell patient data and we do not buy it. What we sell is the infrastructure that lets an institution do something with its own.
Start a conversation
The infrastructure
Built to run where it has to run.
The constraints that sink other deployments are the normal conditions here. They are in the specification, not in the footnotes.
- 01
Offline first
Capture does not stop when the network does. Data moves when the link comes back, without anyone having to do anything.
- 02
Sovereign hosting
The platform runs where the institution needs it to, including on its own servers. Where the data sits is their decision to make.
- 03
API first
Everything the platform does is reachable from the systems already in place. We do not ask anyone to abandon their tools.
The proof
Panzi, Bukavu the first deployment.
Panzi General Referral Hospital has kept continuous clinical records since 1999. We are digitising them, structuring them, and plugging the tools that use them into the result. It is the whole platform, running, in a real institution.
See our work at Panzi