Built in Zimbabwe
We build AI and IoT systems that run where the work happens — on the tank wall, in the field, behind the till — and keep running when the connection does not.
What we build
We are a small engineering company. We take on work we can finish and support, and we write the software ourselves.
Models that run on the device instead of in somebody else's data centre. Vision and sensor inference at the point of capture, so a decision takes milliseconds, costs no bandwidth, and the data never has to leave the site.
Sensors, gateways and the software behind them, specified for one job rather than bought off a shelf and bent to fit. Fish-farm monitoring — dissolved oxygen, temperature, feeding — and agricultural sensing across soil, water and shed.
The systems a company actually runs on, built to fit how it works: point of sale, accounting, human resources, inventory and the reporting that ties them together. One register of the truth instead of five spreadsheets.
Systems
Platforms we have designed, built and run in production — each one started as somebody's awkward, specific problem.
A whole fish farm in one system. Tank-side capture by QR code, dissolved-oxygen and temperature monitoring, feeding by ration and bucket, stock and harvest, an abattoir intake log, stores, inventory, HR and the books — with a daily watchdog that reports feed leaving the store that nothing downstream accounts for.
A till and a back office that keep selling through a dropped line. Offline-first by design, with the queue reconciling itself when the connection returns, multi-store stock, shift accounting and receipts that survive a power cut.
Device management, live datastreams and control for sensors in the field: readings in over MQTT, thresholds and alerts configured per device, and a mobile app to watch and switch things from wherever you are standing.
Android apps for people whose hands are wet and whose signal is not guaranteed. Scan a code, record the reading, carry on — the capture queues on the handset and syncs when there is a line, with photographs and the moment it actually happened attached.
Why the edge
Most business software is written on the quiet assumption of a good line and a nearby data centre. Neither is a safe assumption here. A system that stops when the network does is not a system — it is a liability with a login screen.
So we build the other way around. The work happens on the device: the reading is taken, the sale is rung up, the model draws its conclusion. Synchronising with the server is something that happens afterwards, and nobody has to wait for it.
A decision made beside the sensor does not wait for a round trip to another continent.
Send the conclusion, not the video. Most of what a sensor sees never needs to leave the site.
Capture queues locally and reconciles later, with the time it actually happened intact.
Processing on your own hardware is the simplest privacy guarantee there is.
How we work
We come and watch the job being done. The useful requirements are never the ones in the brief.
One real workflow, end to end, in your hands early — so the design argument happens over something that runs.
Onto the hardware, on the wall, in the hand. Installed and tested in the conditions it has to survive.
Support, changes and the next slice. Software that nobody maintains stops being an asset quickly.
Get in touch
A sensor you cannot read from the office, a stocktake that never balances, a model you want running on a device instead of a bill. Send us the awkward version of the problem — that is the one worth solving.
