With IoT becoming the norm across many industries, manufacturing to retail to public utilities to supply chain to many verticals across the spectrum, it becomes important to focus on the intelligence needed to ensure that this truly distributed system of devices are working seamlessly. In fact, we would even claim that the role of IT goes beyond centralized infrastructure to a perimeter encapsulating all the edge devices. Having intelligence to manage these edge devices becomes critical in any modern enterprise. This is the problem Swim.ai, a new startup in this space, is trying to address.
Intelligence at the edge is not an easy problem. Either you need to spend the data to the central cloud for processing or bring powerful compute near data to learn from large volumes of data. Sending the data to the cloud for processing adds latency, making the term “real time” a joke. Imagine your self driving car having to send data to the central cloud for managing a near death situation. The centralized cloud for processing the edge data becomes completely meaningless in many situations. Then you have security, regulatory and data privacy issues which may prevent some data from going to a central cloud. On the other hand, bringing powerful compute to the edge and learn from large volumes of data is expensive and, in certain cases, inefficient. Using machine learning for intelligence at the edge is both a complex distributed computing problem and a complex machine learning problem.
Swim.ai: A framework for machine learning at the edge
Swim.ai is a new startup that is launching today offering a new framework to gain insights from edge devices without having the complexity that comes with such a distributed architecture. Swim.ai framework is
- Plug and play that can be deployed even on low powered devices like smart meters
- Real time processing with data that is relevant for the localized region. There is no need to manage huge volumes of data or move the data to a central cloud for processing
- Secure as starts with an immutable hardware root of trust to ensure that the system will always bootstrap into a known-good initial state. More importantly it continuously learns about the device and its environment to offer better security
The framework is flexible enough to run on devices at the edge using a mesh like network, train on data present locally on these devices and provide real time insights. Such a flexibility makes Swim.ai an interesting framework to watch as more and more devices are added to the edge, driven by IoT use cases.
Here are some use cases I could imagine for a framework like Swim.ai but there could be more
- Supply chain logistics
- Retail, both on the inventory management and consumer interactions
- Smart cities
- Self driving automotivecs and the logistics with managing fleet
- Traffic management both with or without autonomous vehicles
Swim.ai is an interesting startup, especially from the context of our interest on the role of AI in Observability and, in the future, on Operations. With IoT becoming mainstream in the modern enterprise, a framework like this will play an important role in ensuring better business outcomes.