A well-known problem in the Internet of Things is that many connected devices operate in silos. Your Fitbit doesn’t communicate with your Nest thermostat, for example.
One way some companies are trying to solve this problem is to create a hub, like Revolv. The idea is for all devices to connect to the hub, which serves as a central point for users to control all the devices and allows certain events to trigger activity in different devices.
Neura, a startup chosen as part of Microsoft Ventures’ accelerator program, has a novel approach to the hub concept. “The phone and potentially in the future the watch is how we treat a hub,” said Gilad Meiri, CEO of Neura.
Neura aims to be like a central clearinghouse for IoT data collected from fitness trackers, home automation products, and phones. But then it interprets that data into useful information that it supplies to other devices.
Here’s one scenario. Neura works around the idea of events in a person’s life. An important event could be waking up in the morning. Neura may figure out that a user has woken up based on information from a variety of devices like a sleep sensor, a Nest thermostat, motion sensors in a phone, and historical patterns.
Once Neura has detected such an event, it supplies it to partners that subscribe to that event data. For instance, a TV vendor might want to know a user has woken up in order to turn on the TV to the user’s favorite morning program. Waking up could trigger events like turning on the coffee maker or starting up the hot water heater.
“This is our model. To understand people and events and allow devices and services to subscribe to that,” Meiri said.
Neura offers its business customers a confidence scale around the information it delivers. For instance, the TV app may not want to turn on the TV unless Neura has 100 percent confidence that the user is awake because it wouldn’t be a great user experience if the TV turned on while the user is still asleep. But the app on the hot water heater might instead like to know when Neura is 60 percent sure the end user is awake since it takes some time to heat the water and it might be better to err on starting to heat the water before the user is awake.
Healthcare applications envisioned by Neura get even more interesting. Neura could detect that a user is driving to the gym and predict that in 20 minutes the user’s glucose level is likely to drop, based on historical data collected from the user’s glucose meter during previous workouts at the gym. The service can suggest to the user that it’s a good time to eat an apple.
Neura could also provide information to services so that, for example, a music service like Spotify can get a notification that a user only has 15 minutes left to her run so that the service can start playing music that might motivate her through the final stretch.
On the backend, Neura ingests the sensor data into its translation machine that it calls Harmony. It’s an abstraction layer that normalizes the data that’s coming from different sources. On top of that sits what Neura calls its Trac Event Machine which looks for patterns in user behavior. Its artificial intelligence layer makes sense of the data.