NEURAL NETWORKS.
DEPLOYED WHERE THE GRID ENDS,
THE CLINIC IS DARK,
AND THE HARVEST SPOILS.
Four sectors. Twelve neural systems. Built for the operator economies of East Africa and shipped globally. Trained on local data, named in local languages, deployed where they matter most.
MAKING THE GARDEN SMARTER THAN THE WEATHER.
In Uganda, smallholder farmers lose over 30% of every harvest to pests, spoilage and unpredictable rain. We're building neural systems that put a satellite, an agronomist and an extension officer in every farmer's pocket — in Luganda, Swahili, Runyankole.
MUKENE
A field-vision AI that diagnoses crop disease from a single phone photo.
Cassava brown streak, banana bacterial wilt and fall armyworm wipe out entire seasons. Most farmers can't reach an extension officer for weeks.
- 01Farmer photographs a leaf with a basic Android phone
- 02An on-device vision model identifies disease + severity in under 2 seconds, offline
- 03Recommends treatment in Luganda, Swahili, Runyankole or English
- 04Logs every diagnosis to a regional outbreak map our agronomists monitor
OBULIMI
A satellite + soil-sensor model that tells you exactly when to plant, irrigate and harvest.
Uganda's agriculture is 90% rain-fed. Climate change broke the two-season calendar. Farmers plant on tradition; the rains arrive on physics.
- 01Pulls hyperlocal rainfall, NDVI vegetation index, and soil moisture data per parish
- 02Cross-references 40 years of regional weather patterns
- 03Sends a weekly SMS: "Plant beans now. Skip maize. Heavy rain in 9 days — drain the lower plot."
- 04Free for farmers under 2 hectares, subsidized by carbon offset partners
POSHO
A cold-chain + market-routing AI that cuts post-harvest loss from 30% to under 8%.
A farmer's tomato in Mbarara might rot before it reaches Kampala. The trader doesn't know the price; the buyer doesn't know the supply.
- 01Aggregates real-time price data from 47 East African markets
- 02Predicts spoilage windows by produce type, weather, distance, transport
- 03Routes cooperative truckloads via the highest-margin path
- 04Optional IoT temperature sensors for high-value cargo
THE CLINIC IS DARK. THE NEAREST DOCTOR IS FOUR HOURS AWAY. WE CHANGED THAT.
Uganda's maternal mortality ratio is 189 per 100,000. Rural clinics deliver babies by phone-light. Half the vaccine stock spoils in the cold chain. AI doesn't fix infrastructure — but it can buy back the time and accuracy that infrastructure costs.
MAAMA
A prenatal risk-prediction model that flags high-risk pregnancies six weeks earlier than standard care.
Most maternal deaths in Uganda are preventable — pre-eclampsia, hemorrhage, obstructed labor — but rural midwives don't have ultrasound, lab work, or specialists.
- 01A midwife enters 14 simple data points on a tablet (BP, age, parity, fundal height)
- 02A neural model trained on 200K East African pregnancies returns a risk tier
- 03High-risk cases get auto-referred to the nearest hospital with a transport voucher
- 04Anonymized data improves the model continuously
KIPIMO
A multi-disease screening AI that turns a basic phone camera into a triage tool.
A community health worker in a village of 800 is the only medical professional within 60km. They make 30 decisions a day with no equipment.
- 01Phone camera + flashlight captures eye, skin, throat, blood smear (with attachment)
- 02On-device models give a confidence-ranked differential diagnosis
- 03Connects to a real telehealth doctor for any flagged case
- 04Audio mode: cough analysis for TB screening, trained on 1.4M East African voices
DAWA
A cold-chain + supply-chain AI that cuts vaccine and drug spoilage by half.
Up to 50% of vaccines spoil in rural Uganda due to broken cold chains, theft, and bad routing. Children miss immunizations.
- 01Low-cost IoT temperature loggers on every cold box, transmitting via 2G
- 02A predictive model anticipates fridge failures 48 hours out
- 03Routes urgent re-stock from the nearest functioning facility automatically
- 04Live dashboard for district health officers
THE GRID IS A SUGGESTION. THE SUN IS A PROMISE. WE SCHEDULE BOTH.
60% of East Africa runs off-grid or on flaky grid power. Solar panels are everywhere; intelligence about them is not. Our models turn every panel, battery and meter into a node that knows what it costs, who needs it, and when to fire.
JUA
A solar microgrid orchestration AI that balances generation, storage and load across thousands of village systems.
A village solar microgrid serves a clinic, a school, twenty homes and a kiosk. Demand spikes are unpredictable. Without smart routing, the system browns out.
- 01Each panel + battery + meter reports state every 30 seconds
- 02A reinforcement-learning model optimizes dispatch in real time
- 03Prioritizes critical loads (clinic refrigerator > kiosk lights)
- 04Pay-as-you-go billing built in via mobile money
UMEME
A grid-failure prediction model that tells utilities where the next blackout will happen, before it does.
Uganda's grid loses 17% of power to faults, theft and overload. Outages last hours to days. Repair crews are dispatched reactively.
- 01Ingests substation telemetry, weather, historical fault data, infrastructure age
- 02Forecasts probability of failure per feeder for next 72 hours
- 03Pre-positions repair crews and parts to high-risk substations
- 04Crowdsourced fault reports via WhatsApp bot fill the gaps
BARABARA
A road-infrastructure vision model that finds potholes, washouts and collapsed culverts before they swallow a bus.
Rural roads are maintained reactively. A collapsed culvert can cut off a region for weeks. Inspection by humans covers a fraction of the network.
- 01Dashcam footage from public-transport vehicles is processed nightly
- 02A vision model classifies defect type and severity per GPS point
- 03Generates work orders ranked by traffic + criticality + cost
- 04Open-data layer for districts and donors
YOU CAN'T MANAGE WHAT YOU CAN'T MEASURE. WE MEASURE THE UNMEASURED.
East Africa is on the frontline of climate change but largely invisible in global climate models. We're closing that gap — with sensors, satellites and neural systems that map the forest, the air and the water in real time.
MISITU
A satellite + acoustic AI that detects illegal logging within hours of the first chainsaw.
Uganda has lost 40% of its forest cover in 30 years. Patrols arrive days late. The damage is done.
- 01Satellite imagery (Sentinel-2 + Planet) is processed daily for canopy change
- 02Solar-powered acoustic sensors detect chainsaw and vehicle signatures
- 03Combined model triggers a real-time alert to ranger teams with GPS pin
- 04Carbon credit verification layer for forest-protection finance
HEWA
A hyperlocal air-quality model that maps urban pollution street-by-street using low-cost sensors.
Kampala's air pollution kills more people annually than HIV. Official monitoring stations cover four points in a city of two million.
- 01$30 air sensors mounted on matatus collect data on every route
- 02A spatial interpolation model fills in coverage gaps
- 03Live map for residents, schools, hospitals via WhatsApp + SMS
- 04Open API for researchers and policymakers
MAJI
A water-source quality and availability model that tells communities which borehole is safe today.
Hand-pump boreholes break, run dry, or get contaminated. Communities walk hours to discover this. Waterborne disease follows.
- 01IoT sensors on boreholes monitor flow, pressure, conductivity, turbidity
- 02A predictive model flags failing pumps before they break
- 03WhatsApp/SMS alerts to community managers + auto-dispatch repair teams
- 04District-level dashboards for water authorities
TWELVE SOLUTIONS. ONE STACK.
PyTorch · ONNX · TensorFlow Lite · Whisper · Llama-derivative African language models · Custom CV architectures
Mofano.ai edge runtime · Twilio SMS · WhatsApp Business API · Africa's Talking · USSD fallback
Makerere AI Lab · KodeInstitute engineers · Ministry of Health partnerships · Local cooperative networks
WANT THIS DEPLOYED SOMEWHERE?
Government, NGO, philanthropic funder, operator. If you have a problem in one of these sectors — or a sector we haven't named yet — let's talk.
START A CONVERSATION →