By adaptive - August 14th, 2017

A startup aims to make all homes more intelligent thanks to advanced machine learning. After all, smarter homes promise more energy- and cost-efficient dwellings. But it can be hard to make sense of the current situation and costs if you have older appliances that aren’t yet “connected” to the Internet of Things.

That’s where Chris Micali, co-founder and VP of product for Sense comes into the picture. He discussed how the firm’s product operates and its value proposition with Open Mobile Media’s Robert Gray.
OMM: First of all, describe what Sense does for consumers?
Micali: It’s an energy monitor for your home that lets you see what’s going on in your house, how much energy you’re using, by (each) appliance and whether they’re on or off right now. You can see if the oven’s on, the iron’s on, or when the kids get home.You can see garage door open, TV or video game console go on. We can surface what’s going on in-app and infer a lot about what’s going on in your house.
OMM: That’s a tall order. How does it work?
Micali: We’re a box that goes in the electric panel and measures the amount of electricity that’s coming into your house. We use advanced machine learning and signal processing to know which appliances are on and how much power they’re using.
The box in your electric panel connects to your home Wi-Fi network and accesses through an iOS or Android app. The home screen is a set of bubbles, a real-time display of what’s on, say a TV bubble if it’s on. The size of the bubble corresponds to the amount of power it’s using. That answers questions like, “Did I leave the oven or iron on?” Or, “Are the kids still watching TV?”
It’s a timeline like a Twitter feed showing what you’ve been using, in chronological order.
In other sections you can see energy usage in detail, historical usage.
OMM: What reaction do you get from consumers, especially those with older appliances?
Micali: In most people’s homes there’s usually one surprise. That old dehumidifier or old fridge in the garage, or a broken sump pump that’s running all the time. There’s usually one surprise that people can correct and immediately save money.
OMM: How does Sense fit in with the growing smart home devices and connectivity with digital assistants?
Micali: As far as energy monitoring specifically, there aren’t any that can do what we do today. They can tell you total usage and others claim to break out devices but we’re the only one that can show you by device in real-time in the home.
In the broader smart home space there’s a lot of competition for connected devices--Nest, Philips Hue, other things that provide control into lighting or HVAC.
You might get Nest to control the HVAC, or Philips Hue to control lights remotely but we measure where power comes into the home so we have a view on everything and not just one thing. It’s a unified view of the home.
We will eventually connect into Nest or Hue and see them in Sense and then cut them off in Nest.
OMM: You’ve recently announced that Sense is integrating with Amazon’s Alexa. What functionality does that offer?
Micali:  You can ask Alexa: “How much power am I using?” You can also say Alexa, “Is the dryer done?” Or “Is the oven on?”
“Ask Sense what time did the TV turn off?” Or “What time did the kids go to bed?”
You can ask Alexa about the state of devices in your home now or in the past, or to surface devices (to check their power usage). It gets you access to all the non-smart devices in your home.
I don’t need to replace the dryer with a “smart” one, Sense can let me know if the dryer’s done.
OMM: Can you expand on the machine learning that powers Sense?
Micali: It’s a big pattern recognition exercise. In the power signal, every device has a shape of its power use. The mission of Sense is to detect those patterns and attribute to that same device every time. Over time as we collect more data from your house, we use those examples to train models to detect it on the (Sense) box. Your washing machine in your house we use to train boxes, the more Senses we have online the more examples we have improves the quality and recognition of all houses.
We’re using user input, some things we don’t automatically classify, my sump pump in my basement the first time it went on, Sense knew it was a motor but didn’t know it was a sump pump. So I renamed it sump pump in my app and when another 100 did the same thing Sense can automatically start naming it sump pump. Once it has enough examples it can automatically detect it as a sump pump.
It’s a similar process to the speech recognition world, the more examples you have of somebody saying a word the better job you can do training it.
The more data we have the better Sense can do and the more it learns over time.
The lighting example where you can see in Sense I keep leaving basement lights on--you can see that in the app and get a connected switch so you can turn them off from Sense or in the future Sense may shut them off for you as it recognizes the pattern of you leaving them on.
Sonos speakers, which are always on, we can see those and by integrating with them more directly, can show you the energy impact of having these things on all the time.
The Comcast cable box uses 60 watts, it adds up. With this info we can effect change; find a way to turn off Comcast box so it’s not using energy. 
comments powered by Disqus