Representative End User Clients
Representative Automation Clients
Representative Software Clients
I've known about Predixion Software since my time with the Aberdeen Group. However, yesterday's update from Jamie MacLennan and Steve Stover showed how far the company has come. There's now a distinct focus on the Industrial Internet of Things (IIoT) which puts the company firmly back on my radar.
There are clues to Predixion's industrial focus in its funding history - pretty conventional VC's up until the C funding round in July 2013 when GE Ventures weighed in. That, as the website suggests, is often a deeper commitment than the usual cash and senior management nous. It often involves technology exchange too. To add to this, GE also took part in the recent D round of funding, being joined for the first time by Software AG. Software AG is another long-established vendor that is focusing its recently accumulated product portfolio on Industrial IoT. (I wrote about one of its success stories here.)
The path to Predixion's success here lies in two key factors. First, the company has invested in streamlining the process for building and deploying predictive models. This has been the bane of many predictive modelers life - they would need to spend more time beating their source data into shape and managing models than actually building models and fine tuning them. The company has a patent-pending process for this (Machine Learning Semantic Model™, or MLSM). The output from this process is an MLSM package which holds the entire predictive model, self-contained.
And this is the second key factor - that MLSM package is so small it can execute in a Java runtime environment with as little as 1MB of memory. So, for example, Predixion has run models on a Raspberry PI, executing several hundred predictions per second. That makes it ideal for placing it at the edge of the Industrial Internet of Things, providing predictive capabilities at the device, not just in a remote data center. This is going to be critical - the Industrial IoT is going to generate so much data so quickly. We need to move analytics into the incoming data stream in order to pick up on time-critical events - such as impending machine failure or quality issues - ASAP. A small footprint makes that possible. As further evidence of its acceptance in this role, Predixion recently announced that it will be embedded in the next generation Intel IoT gateway.
There are clues to Predixion's industrial focus in its funding history - pretty conventional VC's up until the C funding round in July 2013 when GE Ventures weighed in. That, as the website suggests, is often a deeper commitment than the usual cash and senior management nous. It often involves technology exchange too. To add to this, GE also took part in the recent D round of funding, being joined for the first time by Software AG. Software AG is another long-established vendor that is focusing its recently accumulated product portfolio on Industrial IoT. (I wrote about one of its success stories here.)
The path to Predixion's success here lies in two key factors. First, the company has invested in streamlining the process for building and deploying predictive models. This has been the bane of many predictive modelers life - they would need to spend more time beating their source data into shape and managing models than actually building models and fine tuning them. The company has a patent-pending process for this (Machine Learning Semantic Model™, or MLSM). The output from this process is an MLSM package which holds the entire predictive model, self-contained.
And this is the second key factor - that MLSM package is so small it can execute in a Java runtime environment with as little as 1MB of memory. So, for example, Predixion has run models on a Raspberry PI, executing several hundred predictions per second. That makes it ideal for placing it at the edge of the Industrial Internet of Things, providing predictive capabilities at the device, not just in a remote data center. This is going to be critical - the Industrial IoT is going to generate so much data so quickly. We need to move analytics into the incoming data stream in order to pick up on time-critical events - such as impending machine failure or quality issues - ASAP. A small footprint makes that possible. As further evidence of its acceptance in this role, Predixion recently announced that it will be embedded in the next generation Intel IoT gateway.