RxScanner on Small AI: Offline neural network catches counterfeit drugs in Africa

In 2019, in Cape Town, the Nigerian startup RxAll's AI scanner for counterfeit drugs RxScanner failed due to connection lag with the cloud server in the USA — after the failure, the company's engineers created an offline version of the model for a smartphone in just 2 hours, which gave rise to the Small AI direction.

How RxScanner works

RxScanner is a portable spectrometer that scans a tablet with infrared light and sends its molecular profile to an AI model with a pharmaceutical database. Within seconds, the system identifies the drug or reports that it is counterfeit. The device already worked in pharmacies in more than a dozen countries, including Ghana, Kenya, Myanmar and Nigeria, where counterfeiting of medicines kills thousands of people every year.

Why cloud AI let down at the demonstration

That morning at a hotel in Cape Town, South Africa, the system did not respond. The spectrometer was connected to a cloud model, but the data center was 14,000 kilometers away from the site, and bandwidth was insufficient.

"Our server was in the USA, and just getting the result of one scan took more than five minutes," says

RxAll founder Adebayo Alonge.

Alonge immediately instructed engineers to compress the model into a compact, low-power version capable of working without internet. The finished solution saved the demonstration just two hours later.

• The demonstration failed in 2019 at a hotel in Cape Town, South Africa
• The data center with the cloud model was 14,000 km away from the demonstration site, in the USA
• One scan took more than 5 minutes due to insufficient bandwidth
• RxAll engineers assembled the offline version of the model in 2 hours
• RxScanner is already being used in pharmacies in more than 12 countries, including Ghana, Kenya, Myanmar and Nigeria

What the transition to a compact model gave

The demonstration failure prompted RxAll to create a new version of the device that checks tablet authenticity entirely without broadband internet, computers, and even stable electricity. Alonge became one of the main proponents of the Small AI approach — compact models that work locally on low-power devices instead of cloud data centers.

Small AI is the opposite of the giant language models of rich countries with their hyperscale data centers, billion-dollar investments and disputes about AI consciousness. But for millions of people in developing countries, it is precisely such — small and offline — AI that remains the only available option.

The IEEE Spectrum material describes the Small AI movement as a response to the reality of most of the planet: millions of people live where there is neither broadband internet nor reliable electricity nor access to hyperscale data centers — and for them the only working AI is the one that fits in a pocket.

What this means

The RxScanner case shows: while the industry argues about trillions of parameters and hyperscaling data centers, for healthcare in regions without stable connectivity and electricity, the decisive factor is not the power of the model, but its ability to work autonomously on an ordinary smartphone.

Frequently asked questions

What is RxScanner?

RxScanner is a handheld spectrometer from Nigerian startup RxAll that scans a tablet with infrared light and with the help of an AI model with a pharmaceutical database determines in seconds whether it is a real medicine or counterfeit. The device already works in pharmacies in more than 12 countries, including Ghana, Kenya, Myanmar and Nigeria.

Why was an offline version of the model needed?

Because the cloud version depended on a server in the USA located 14,000 km from the place of use — with weak connectivity, one scan took more than 5 minutes. The offline model, which RxAll engineers assembled in 2 hours, works directly on the smartphone without internet connection.