In January 2022, Amazon announced the launch of the Ring Alarm Glass Break Sensor, which uses artificial intelligence to monitor glass windows and doors for break-in attempts. Ring Alarm Glass Break Sensor detects the sound of glass breaking from up to 25 feet away. If the sensor detects the sound of a glass breaking, it sends a notification to customers via the Ring app.
This announcement was made possible in part due to innovations by Syntiant. The company pioneers new developments in the field of edge computing – technology that leverages machine learning from the cloud for on-device processing across a wide range of devices. Running machine learning on device results in improved battery life and responsiveness, and makes devices like the Ring Alarm Glass Break Sensor possible.
“If your battery-run device is on alert for an event to happen, while sending information to the cloud to process, it burns through the battery quickly,” says Poupak Khodabandeh, vice president of product for Syntiant. “In addition, customers shouldn’t be limited to placing the device near an outlet for connectivity. They should be able to place their device where it makes the most sense—for aesthetical considerations or even near a window.”
The Amazon Alexa Fund was one of the first investors in Syntiant. Khodabandeh views the partnership as a no-brainer for shaping the next generation of ambient experiences. Launched in 2015, the Alexa Fund provides up to $200 million in venture capital funding to startups focused on ambient intelligence, smart consumer electronics, and new media.
Khodabandeh says that the key to designing customer experiences of the future is dependent on making the interface between the human and digital world as intuitive and responsive as possible. To meet this need, Syntiant produces deep neural network processors in microchips that bring artificial intelligence to the edge, while at the same time taking advantage of the capabilities of Alexa.
Syntiant was founded in 2017. Initially conceptualized around it’s unique strength in semiconductor design, Syntiant supplies integrated circuits and microchips for edge applications. In addition, the company also provides use-case specific machine learning models that run on them.
“We’re a hardware and software company, providing everything our customers need to build intelligence into their edge device and get to production quickly,” says Khodabandeh,. The company has seen a growing demand for their integrated offerings. Today, Syntiant has grown to 100 employees across the United States, Asia, and Europe.
“Syntiant was founded on the premise of promoting voice interfaces but has since expanded its offering to include all audio and vision applications,” says Khodabandeh. “In the ’70s and ’80s, we had keyboards to interface with machines. Then came the mouse, which in turn was followed by touch screens. Today, people are increasingly turning to natural interfaces, like voice or gestures, where they interact with technology as naturally as they do with other humans.”
Using edge computing to improve the customer experience
Edge computing is a distributed computing framework that brings applications closer to data sources such as Internet of Things devices or local edge servers. At the heart of edge computing is a philosophy of putting as much computational power as possible on the interface of the digital and human world.
“In our company, our vision is to make everyday technology Syntiant – to state the obvious, our name is a play on ‘sentient’,” says Khodabandeh. “We want to get to a point where devices interact seamlessly with humans: they can listen, see, feel and make decisions using information on the edge. We want to eliminate the massive congestion of data between devices and cloud servers.”
The startup has developed several technical innovations to facilitate edge computing. Syntiant’s neural decision processor mimics the human brain for running massive workloads efficiently. The silicon architecture combined with an edge-optimized training pipeline and data platform enable edge processing of machine learning models.
For the Ring Alarm Glass Break Sensor, Syntiant used real-world data with a wide variety of synthetic background noises to train the machine learning models to discern between sounds like the breaking of a window, to other sounds such as a glass shattering on a sink.
“Our technology runs that detection mechanism on the device and only sends the information to the cloud when it’s certain that the event has occurred. So customers aren’t falsely alerted and don’t needlessly burn up the battery,” says Khodabandeh.
Running the detection mechanism on the edge also protects the privacy of customers because it only sends data related to a specific sound, not all sounds in the home. “You don’t want non-relevant, private data sent to the cloud,” explains Khodabandeh. “When devices process the data at the point that it’s generated, while only transmitting the end decision to the cloud, we ensure that customer privacy is protected.”
Growing with Alexa
Syntiant was connected to the Ring opportunity through the Alexa Fund. Khodabandeh says that a fifth of the startup’s business to date has been secured because of the strong technology integration with Alexa.
“We are able to go to tier one accounts, and enable them to implement an Alexa solution on their device,” says Khodabandeh. “Without the relationships that we built through the Alexa Fund, we wouldn’t have been able to experience such rapid growth.”
The pervasiveness of Alexa means its specifications and standards are well understood and defined, which streamlines the process of testing, verifying, and qualifying Syntiant solutions.
“When you are a startup, first shipments matter. The quicker you get to those first revenues, the more stability the company has. Alexa is used on a lot of consumer devices, so it opens up a lot of opportunities,” explains Khodabandeh. “We want our consumer applications to have the Works with Alexa certifications, because then they get taken seriously.”