General Electric’s Predix platform for the industrial Internet of Things (IoT) has its brainpower in the cloud, but the company has plans to replicate those smarts at the network edge.

The enhancements, which GE hopes to complete during the next four to six months, would let developers take advantage of the Predix platform-as-a-service (PaaS) without always having to contact the cloud.

The plan reflects how urgently GE wants to build a developer community around the PaaS. Predix was launched to the public five months ago, and already, GE has held its first-ever Predix conference — Predix Transform, which concluded yesterday in Las Vegas. Now it wants to invite developers to not only run Predix on the things side of IoT but to add their own algorithms and software as well.

"What we're trying to enable is this capability to write applications that can do something with the data at the edge," says Hima Mukkamala, head of engineering for Predix.

That's the key — doing something. Predix's end goal is to improve some aspect of the physical asset it's watching, be it a jet engine or a factory robot. In the cloud, Predix can crunch the math on a computer model of the asset (a "digital twin," GE likes to call it) and send the asset new instructions based on the results.

Not all of these assets can be in continual communication with the cloud, however; one common IoT assumption is that some assets will preserve battery life by talking to the network only intermittently. Moreover, you wouldn't always want to wait for the asset to receive instructions from the cloud. A wind farm, for example, could reap big rewards from slight, real-time alterations in the way individual turbines run.

"Those are the kind of things that we're looking for: small tweaks that take an already efficient system and make it even more efficient. Because the industrial environment is at scale, you get these returns that are kind of crazy," says Greg Petroff, GE's chief experience officer.

So, GE wants to move more of Predix's smarts to the edge. In many cases, this could mean running the digital twin right at the edge, with local software making immediate corrections to the physical asset.

And yes, there's a container angle.

Devices at the edge are holding more and more CPU power, giving them the headroom to run more applications. GE wants to exploit that by moving Predix to a container-based architecture. Containers, in general, are meant to make it easier to move applications from one environment (such as Predix in the cloud) to another (such as Predix at the physical asset).

"The seamless movement between the cloud and the edge — that's where we're getting to in the next four to six months," Mukkamala says.

Finally, GE wants to use APIs to let developers add their own algorithms and applications into Predix, either on the edge or at the cloud. That way, developers could give Predix's algorithms an industry-specific spin — or even a location-specific one.

"An airline that flies out of Dubai is going to have a very different profile from one that flies out of San Francisco, just because there's sand in the first 5,000 feet of air. Knowing that is going to change the behavior of how you're going to manage your fleet," Petroff says.

Photo: Naresh Mehta (@nkumehta) on Twitter, taken during Predix Transform.