a greenhouse with connected cameras and sensors

AWS Greengrass runs ML models on the cameras monitoring plant growth, noting the changes in conditions, allowing for corrective actions to be made. Source: Yanmar

Growing with Greengrass: running analytics where you need it

TODAY, businesses deploy IoT systems to collect information and gain key insights about their company, its products, and its end users.

Current IoT systems rely on network connections to function, which is why businesses struggle to deploy them in remote areas with limited coverage.

Tech companies’ solution to this, is to run analytics natively where the edge devices are. This is sometimes known as the intelligent edge. Amazon’s Greengrass does that by allowing edge devices to run machine learning (ML) algorithms on the sensors itself.

Machine learning algorithms are used to learn patterns and identify anomalies in data. For an IoT device, this means it doesn’t need to be constantly connected. The device itself can recognize what needs to be done, and send alerts to the cloud only when actions needed to be taken.

This significantly reduces network usage, which helps drive costs down as well. The costs of bandwidths aren’t cheap to begin with.

Not every IoT device is able to support a full ML algorithm.  How Greengrass works, is to only run the ML model on the edge device. The bulk of the learning is still done in the cloud through AWS IoT Analytics. The device is then periodically updated when upgrades are available.

Many companies across Asia are already adopting AWS Greengrass, for different functions. Michael Garcia, Senior Technical Program Manager of AWS IoT, shared with Tech Wire Asia some examples in a recent media sharing session.

Diagram of how Yanmar’s system works. Source: AWS


Japanese company Yanmar specializes in diesel engine design and manufacturing. With a huge presence in smart farming, it installed cameras to monitor plants growth and predict yields.

By deploying ML models on the greenhouse camera ecosystem, the cameras could process plant images natively. When it detects anomalies, it triggers alerts and initiates corrective steps to help the crops grow better.

The cameras previously sent huge amounts of image data to the cloud, which takes up a lot of bandwidth. Yanmar has significantly reduced its 3G network costs with ML.


Singapore based Zimplistic has a slightly different approach. It uses intelligent edge to improve customer experiences.

Zimplistic developed a flatbread making device called Rotimatic; a connected smart device that makes fresh roti from scratch.

The company runs AWS Greengrass to monitor performance of the machines. Zimplistic gets notified when errors occur, so it can make changes to the software and push updates over air immediately. It also gathers data on customer usage, which is then fed back to design updates.

For consumers, this translates to a smart device that can repair itself. This shows in its successful sales globally, recording 20,000 devices sold in 12 months.

Endless possibilities

These are just a couple of examples in Asia. Globally, AWS works with many other companies and industries, including Phillips in healthcare, Denso in automotive, British Gas in utilities etc.

Every company will use IoT differently. Companies need to determine how they want to use real-time insights and data.

Bottom line, it’s about focussing on business outcome and strategy, improving efficiency, and building a better relationship with your customers.