Pawanax Research
Clinical AI, African languages, epidemic intelligence, forecasting and the economics of health. We publish our methods, our results and our forecasts, and we score those forecasts in public.
Pawanax Research is the research group of Pawanax AI. We study clinical AI and safety, language intelligence for more than 50 African languages, epidemic intelligence through Epicenter surveillance, forecasting and decision science, health economics including remittance flows, and critical minerals and industrial policy. The group is led by Dr. Fidelis Ndombera, PhD, whose six peer-reviewed publications cover cancer biology, medicinal chemistry and COVID-19 cheminformatics. We publish methods and results, not proprietary model internals, and we register forecasts in advance and score them with the Brier score.
Live lab instruments
As of
internship & academy applications
1,755 in the last 90 days
WhatsApp agent messages handled
across 167 user profiles
epidemic-surveillance alerts generated
from 24 monitored data sources
registered platform accounts
plus 82 stablecoin wallets
Research areas
Decision support for clinicians and families, evaluated per task on quality, safety and cost. Pawanax supports care; it does not replace a doctor, and urgent symptoms are escalated to human care.
Questions we are asking
Understanding and generation across more than 50 African languages, so that people can ask about their health in the language they think in.
Questions we are asking
Epicenter surveillance turns scattered public and local signals into alerts for public-health teams. So far it has generated 5,329 alerts from 24 monitored data sources.
Questions we are asking
Probabilities stated in advance and scored when they resolve. We track calibration, not confidence, and we publish the forecasts we get wrong.
Questions we are asking
How money moves into and out of health. Kenya's formal remittances reached $5.04B in 2025, and 12.5% of that value goes to medical care.
Questions we are asking
How mineral value is captured locally or exported raw, and what industrial policy can do about it. Work in this area runs through Matawis.
Questions we are asking
Library
Working papers, theses, forecasts and protocols from the lab. Each PDF is the version we stand behind, with its date and version noted on the card.
Featured
A scenario forecast across economics, geopolitics, technology, science and psychology at 3, 6 and 9 months and 1, 2 and 3 years
A reproducible scenario model of the world as of 13 October 2026. It separates observed, preliminary, forecast and judgment-based evidence, simulates five pathways with correlated monthly shocks (100,000 draws), registers dated hazard probabilities, and translates the result into a ranked register of 40 commercial opportunities for Kenya and East Africa. Its limits are stated plainly: the model is reproducible but not yet calibrated, and it proposes a Brier-score protocol to earn an accuracy claim.
3 papers
An operating thesis for turning abundant machine intelligence into Kenyan-owned physical prosperity
A 56-page founder thesis on how Kenya can convert increasingly capable and inexpensive machine intelligence into durable national prosperity across energy, agrifood, cyber trust, programmable settlement and regional infrastructure, with measurable dashboards for debt service, reliable power, productivity, export complexity, household welfare and domestic capital ownership.
A decision framework with reproducible simulation diagnostics and a prospective Kenyan fluorspar validation study
A methods and simulation study proposing how investors and governments can weigh evidence that minerals were actually processed where claimed. It combines three evidence layers across 20,000 simulated paths and five conditions, reports failure modes rather than hiding them inside a single accuracy score, and pre-specifies a prospective field validation on Kenyan fluorspar. Data are synthetic; field validation and buyer adoption remain prospective.
Clinical intelligence, WhatsApp access, disease surveillance and instant settlement on one platform
A short whitepaper describing how Pawanax connects clinicians, patients, surveillance and payments: phone-native access with zero install, clinician-facing intelligence, outbreak signal monitoring and a settlement rail, designed around the way healthcare already works in Africa.
Peer-reviewed work by our team
Journal of Pharmacy and Pharmacology · 2019
F Ndombera, G Maiyoh, V Tuei
Reports the pharmacokinetic, physicochemical and medicinal-property profile of an n-glycoside anti-cancer agent, and compares its activity with 2-deoxy-d-glucose in lung cancer cell lines.
51citations
Google ScholarMolecular BioSystems · 2016
KTG Samarasinghe, DNPM Godage, Y Zhou, FT Ndombera, E Weerapana, YH Ahn
Describes a clickable glutathione approach for identifying protein glutathionylation in cells responding to changes in glucose metabolism.
40citations
Google ScholarBioorganic & Medicinal Chemistry Letters · 2016
FT Ndombera, GC VanHecke, S Nagi, YH Ahn
Examines carbohydrate-based compounds designed to induce cellular stress as a strategy for targeting cancer cells.
10citations
Google ScholarPure and Applied Chemistry · 2017
FT Ndombera
A review of anti-cancer agents and reactive oxygen species modulators that act on altered metabolic pathways in cancer cells.
8citations
Google ScholarJournal of Bioinformatics and Computational Genomics · 2020
FT Ndombera
A cheminformatics re-examination of the proposed mechanisms of action of chloroquine and hydroxychloroquine in the context of COVID-19.
4citations
Google ScholarCancer Causes & Control · 2022
VC Tuei, GK Maiyoh, FT Ndombera
An epidemiological study of the contribution of infectious agents to cancer causation in Kenya.
2citations
Google ScholarMethod
Every piece of work follows the same five steps, so a claim can be traced from question to score.
01
A decision someone has to make, written down before any analysis begins.
02
Sources, dates and uncertainty recorded line by line, so every claim can be traced.
03
Models are built and judged per task on quality, safety and cost, with a clinician or analyst in the loop.
04
Probabilities are published before the outcome is known.
05
Once a forecast resolves, it is scored with the Brier score, whether it flatters us or not.
University of NairobiKenya
Kenyatta UniversityKenya
Kibabii UniversityKenya
KEMRIKenya
Rice UniversityUSA
Questions
Pawanax Research is the research group of Pawanax AI. We study clinical AI and safety, African language intelligence, epidemic intelligence, forecasting, and the economics of health. The group is led by Dr. Fidelis Ndombera, PhD.
Yes. Every paper in the library above can be downloaded as a PDF. Drafts that have not been published are not available.
We share methods and results: how we built and evaluated a system, and what it found. Proprietary models and production code remain private.
Write to pawanax@pawanalife.com with the subject “Research collaboration” and describe the question you are working on. We reply to every serious proposal.
Our research supports clinicians and families. It does not replace a doctor, and urgent symptoms are escalated to human care.
Researchers, funders, public-health teams and journalists are welcome. Tell us what you are trying to find out.
Start a research collaboration