Everything “home” is turning “smart” these days. So how about a wall outlet?
CareAlert is taking simple wall outlets that power practically everything in your home and turning them into an AI-based emergency beacon. Grit Daily caught up with CEO Fereydoun Taslimi to unpack what these latest devices have in store for us.
- Your team has its own interesting tech background. Share that.
I attended the University of London where I applied for several patents, one was what I believed to be the world’s smallest chair that would fit in your shirt pocket. After completing my studies in London, while attending Georgia Tech for Graduate work, I came up with the idea of a multilingual computer, this was pre-PC time. I worked with a local terminal maker and turned it into a computer and started selling it in the Middle East.
In Early 1980 AT&T was broken up into smaller regional operating companies and the business telephone market was opening up. I saw an opportunity to develop a smart switch. After a year, my team and I had a product (Adax) that would turn any dumb telephone into a smart business phone. It had text to speech and lots of features.
We almost secured VC funding for manufacturing with Ameritech (now an AT&T Holdings company) but the investors pulled out last minute as the AT&T breakup agreement would not allow them to get into manufacturing. For the first time I had cold sweat running down my body as our team had been told to ramp up production as money was coming. After facing some difficulty, I abandoned the hardware, pulled some of the software out of the project and got into voice processing, fax servers and interactive voice response systems (enhanced systems), later my company merged with another company and went public.
After this venture I formed another company (Penumbera), developing Java development tools in 1996. That company did not go anywhere as Microsoft and Sun and others started giving away their tools for free. I built several other companies in the area of information discovery, one called Infogo with the exact functionality of Dropbox in 2000 but at the time getting an application with cell phone carriers was near impossible. Later I got into infrastructure monitoring and managed service.
Two and a half years ago in2015, I saw the struggle my wife was going through to look after her mom and that gave me the idea for CareAlert. Of course, having been away from hardware development for many years it did not occur to me that the development may take much longer than it used to.
The other founder of SensorsCall is Lili Varzi, my wife. She is a technologist and also attended Georgia Tech specializing in SAP and CRM. Other team members include Anand, who has 15 years of experience working with IBM at the highest level. He is a managing partner at an out-sourcing company and also a cofounder of renewable energy company Ztric. What makes him different is the fact that he still likes to code and dive into development. Mitch Kapa has many years of experience designing hardware for consumer electronics and specializes in making them cost effective and work out of the box. Payman Arabshahi was a research scientist at Jet Propulsion Laboratory NASA and has published hundreds of scientific papers.
- How does CareAlert work?
You place between 2-5 of the CareAlert devices that act and look like an emergency light in the house, one in the bathroom, one in the bedroom and another in the living room near the kitchen. They simply plug into your wall outlets.
It takes about one week for the system to learn your habits like what time you go to bed and how long is the shower you take. Once it learns the habits of the household, anything out of range of these “norms” will cause the device to generate an alert. In the first months you can mark some of the alerts as normal until the system stabilizes and learns what is truly abnormal.
- What spurred its development? Any wild story?
It was 11 p.m. when my wife Lili and I anxiously jumped in the car. Lili’s mother had gone out shopping earlier that evening and should have made it home by then, but we had yet to hear from her. Her mother’s cell phone was going straight to voicemail and there was a busy signal on the landline, causing Lili to be concerned.
We jumped in the car and the three-mile drive to her mother’s house seemed like thirty.
By the time we arrived we were sure something was wrong. We rushed inside only to find her mother sitting contently in the living room playing solitaire on her iPad. Her landline had been accidentally knocked off the hook and her cellphone were in her purse, having run out of battery before she got home from shopping. Everything was fine.
On the drive home that night my relief quickly turned into curiosity. Since she is elderly and lives alone, there must be some way to make sure she’s safe without resorting to frantic phone calls and late-night drives.
My first, and in hindsight obviously missguided, solution was to install cameras in her house, allowing us to look in whenever we wanted to make sure she was OK. This was met with understandable resistance and quickly dismissed as a viable option (as anyone who has read 1984 would agree). So, back to the drawing board I went.
I soon realized that I should not be focusing on simply monitoring a person but monitoring their environment and how they interact with and affect it. By taking this approach, I realized I could design a device that would ensure the safety and security of Lili’s mother without compromising her privacy.
After several months of trial and error, the first CareAlert prototype came to life. It was a self-contained, battery-operated device that would be placed throughout the home to monitor a variety of environmental conditions. These devices would include multiple sensors pre-set to certain thresholds and would sound an alert if those thresholds were met or exceeded.
This initial device; however, had its limitations. The power circuit was designed so the battery would last a year, but after testing, it only lasted one month. Moreover, the thresholds had to be set on the sensors, which greatly reduced the ability of the device to adapt to different baseline conditions.
At this point I decided to reimagine the entire architecture of the device. Artificial intelligence and deep learning had been advancing at a rapid pace, and I wanted to incorporate those technologies. CareAlert had to make sense of the information it was collecting and not simply provide the end-user with data points. It had to learn patterns, predict outcomes and provide actionable data to the end-user. After discussing the idea amongst a few engineer friends with a wealth of experience in hardware and software design, I decided to put a world-class team together and embark on developing the most advanced, intelligent, software-driven environmental monitoring device available.
Several iterations later, the team decided on a design similar to that of a night light. It is compact in size and has a removable power adaptor, allowing it to be placed in a variety of spaces and still be connected to power via a USB cable.
To create a “living baseline”, its sensors collect an initial week’s worth of data about the user’s environment and send that data to the cloud for analysis via our proprietary algorithms. Once the device has learned the initial baseline, it can monitor an environment and relay any anomalies back to the end-user via a mobile app.
And recalling my wife’s inability to reach her mother by cellphone or landline, we integrated a speaker though which the end user can speak to her loved one via the CareAlert mobile app. This feature can also be used for other purposes for example, to leave a timed message reminding a loved one to take her medication.
What began as a way to ensure Lili’s mother was home safe became a multi-year project aimed at creating an intelligent, interactive monitoring system that ensures privacy and provides peace of mind.
- What is “normal” household activity?
This varies by individual. That is why the devices take a week to learn a person’s habits and recognize her daily patterns. Examples of what would be flagged as abnormal could include something like, if your shower typically takes 10 minutes and one day it takes 30 minutes that is abnormal, or if normally you go to the bathroom once a night and one night you go five times, that is abnormal. Normal would be what a person typically does every day, but there is no set standard defining normality.
- How does CareAlert learn?
The CareAlert devices collect data and interpret them into actions and assign a normalcy value to them depending on what has been learned in the past. At this time the learning is done in the cloud. The devices do not watch or listen to words.
The more data CareAlert devices collect data, we should be able to detect loneliness, social interaction, possibly behaviors that could warn of impending dangers, that is a goal.