top of page

Shoreline IoT Has Built A Programmable Industrial Sensor



Stacey Higginbotham (Stacey On IoT)

The Internet of Things Podcast - Stacey On IoT

Tech Journalist


Stacey Higginbotham


Building sensors for enterprise or industrial IoT deployments is tough. But an 8-year-old startup thinks it has found the keys to successfully doing so by rethinking business models, identifying the true technical challenges, and taking advantage of the cloud. Kishore Manghnani and Mark Stubbs are the co-founders of Shoreline IoT, a company that makes ruggedized sensors that gather data, send that data to the cloud for analysis, and then provide a SaaS service. When Manghnani, who is also the CEO of Shoreline and was previously at Marvell, and Stubbs, who is also the company's CTO and was previously at Google, started a company aimed at building sensors combined with machine learning (ML), they figured the ML models would be the tough part. But the real challenge was building the device and the related infrastructure needed to support businesses using those ML models.


Shoreline IoT Has Built A Programmable Industrial Sensor

Shoreline's Smart Sensor collects data, but the value is in the analysis it provides. Image courtesy of Shoreline IoT.


The result is a ruggedized module containing six individual sensors, what Shoreline calls a Smart Sensor. The Smart Sensor can be programmed to alert based on several different use cases, from tracking vibrations on a motor for predictive maintenance to measuring methane leaks on pipelines, as noted in the story above.


Each device gets connected to the cloud via a cellular connection. Shoreline chose the cellular connection option because it's more reliable in remote locations and because it's easier for a technician to set up a cellular device than it is to provision a device onto a Wi-Fi network. The Smart Sensors are provisioned in the cloud, and the customer chooses the particular models he or she wants to run on each device.


The sensors are then given to technicians who can scan them and place them in locations as needed. Later, they can be removed, reprogrammed, and placed in other areas. The key to Shoreline's success so far is that these Smart Sensors are so easy to deploy; they can also survive in harsh environments.


They do, however, have to maintain a connection to the cloud in order to work. Indeed, despite presenting at the TinyML summit and showing off models running on Tensor Flow Lite, the big value is in sending the data up to the cloud for analysis, and alerting clients in a cloud-based dashboard. That's not to say that some alerts might happen locally, but in the case of monitoring harsh or remote environments, even an alert delivered on the spot through a local model would be like a tree that falls in the forest when no one is there to hear it.


Which is why the subscription model is so important for Shoreline. The cloud connection to the sensors ensures that the analysis gets communicated back to the appropriate person, wherever they may be in the world. And the subscription model ensures that customers pay for the actual cost of the device and the running of a cloud service (complete with necessary data storage and compute for training new models). It also means customers can get new hardware as needed, such as when Shoreline adds new sensors or develops new products.

Shoreline is just one of many enterprise IoT companies using a subscription model to provide the popular combination of sensors and analytics. Others include Density, which sells a subscription for people-counting; Vutility, which sells a subscription for power tracking on utility lines; and Augury, which sells a subscription for machine health built on its custom sensors and models.


This means that giant companies may soon find themselves awash in custom sensors that are easily deployed and come with a monthly fee. But for this stage of the enterprise IoT, it's a model that makes sense. Most importantly, it recognizes that the value isn't in the sensing, but in the insights.

75 views
bottom of page