In rural areas, infrastructure like hand-operated pumps can greatly improve living standards and drive economic growth, especially in developing countries. However, maintaining a reliable infrastructure requires proper installation and upkeep, which is often neglected due to under investment.

System failures and downtime are more common in rural areas due to limited access to spare parts and skilled labour. In sub-Saharan Africa, up to one third of the hand-operated pumps used daily by 200 million people are often broken and remain un-repaired for a month.

Current remote health monitoring systems for rural infrastructure lack advanced analysis and insights due to limited resources such as battery life and data transmission. Basic data loggers are used, but they lack real-time insights on system changes.

However, an invention by Oxford University has proven that lightweight machine learning approaches can perform anomaly detection within the embedded system despite limited resources. This can be combined with cloud-based machine learning algorithms for distributed inference, providing robust information without the need for expensive hardware or sensors, making continent-wide remote monitoring systems feasible.

The machine learning technology, originally intended for monitoring hand-operated water pumps in rural areas, can also be applied to other types of sensors used for infrastructure monitoring.