The Best Things Come in Small Packages

There’s an old saying, “The best things come in small packages.” The idea is that it is the smallest gifts — such as jewelry or ornaments — that are actually the most precious and valuable. The same can be said about Elliptic Labs’ Small Sensing Models, our small, proprietary Deep Neural Networks at the core of our AI Virtual Smart Sensor Platform.

With the recent news from OpenAI about their impressive hockey stick growth to Apple’s introduction of Apple Intelligence, the market is ready and eager to see how AI is truly adding value and transforming our future. Therefore, we are eager to share Elliptic Labs’ AI bona fides and how our AI innovations have helped to shape smartphone industrial design and now bring exciting and critical user experiences to the PC space.

We have been calling our 100% software-only products “AI Virtual Smart Sensors” since 2018. Today I’ll explain why. First, I am going to introduce Elliptic Labs Small Sensing Models, sharing some details about our AI innovations. Elliptic Labs has been developing these Deep Neural Networks since 2016, we launched to address a few issues for our first smartphone customer.

In 2016, we launched our first product with Xiaomi on the game-changing Xiaomi Mix smartphone, the first full-screen smartphone design ever. Our INNER BEAUTY? software product (now called the AI Virtual Proximity Sensor? INNER BEAUTY) was at the core of that design. Because we were able to eliminate the need for a physical infrared (IR) sensor to do the proximity sensing, there were no more unsightly holes and instead a cleaner aesthetic with thinner bezels. Our product paved the way to an elegant, futuristic full screen design

Issues needed to be address

1)??????? Scalability – how would we be able to develop and support the many smartphone models and high volumes Xiaomi was expected to ship?

2)??????? Performance and reliability – how would we be able to reach and surpass the 99.9% accuracy of the existing hardware component sensors?

However, this first launch, though an impressive feat of engineering, was fraught with some issues. AI Virtual Proximity Sensor INNER BEAUTY was an elegant, technically beautiful solution that combined our (ultrasound) signal processing prowess and advanced know-how to integrate into the complex world of Qualcomm and Android architectures. Our ultrasound expertise enabled us to (almost literally) hand tune our algorithms without the aid of automation and scalability (read: AI and machine learning). I’m sure you can see the downside of this solution, with the introduction of new corner-cases and environments requiring more and more manual hand-tuning. This resulted in a parade of Elliptic Labs engineers flying to and from Beijing weeks at a time to ensure the success of the product launch – both ours and Xiaomi’s.

Despite these issues, the launch was an unmitigated success, rocketing Xiaomi into the mainstream consciousness and placing their brand amongst the likes of Apple and Samsung. To this day, Xiaomi is a top-3 smartphone maker and still a big Elliptic Labs customer.

Unfortunately, this approach was not scalable. As Xiaomi grew to become the behemoth it is today, we knew we were not going to be able to continue throwing over-qualified engineers and excessive people-hours at each subsequent product launch. We needed to scale and to ensure superior reliability and robustness. This is when we introduced the Elliptic Labs’ AI Virtual Smart Sensor Platform.

How Elliptic Labs’ AI Virtual Smart Sensor Platform solved these issues

1)??????? Significantly decreased time to market and required resources by utilizing AI and machine learning to automate training, testing, and delivery processes of our AI Virtual Smart Sensors.

2)??????? The continued accumulation of datasets allowed us to solve for more corner-cases rapidly and successfully, increasing performance and robustness to surpass hardware sensor component performance

The core of our AI Virtual Smart Sensor Platform is our Small Sensing Models — deep neural networks that we deliver to customers as fully integrated AI Virtual Smart Sensors. In addition to developing the platform, we also developed an automated process for training, testing, and tuning these AI Virtual Smart Sensors, eliminating the burden of manual work.

Thanks to our foresight in creating this platform, by 2018 we were fully equipped to address the expansive and mature smartphone market, introducing our AI Virtual Proximity Sensors not just to Xiaomi, but to the smartphone industry at large. What started as a significant manual engineering effort had become a successful, stable, and scalable solution

Today, Elliptic Labs has shipped its proximity solution on over 130 smartphone models from 11 different smartphone makers, placing AI Virtual Proximity Sensors on over 500 million devices. We are now repeating the success of our AI Virtual Smart Sensor Platform in the PC space, with great results. At the center of all of these developments has been our Small Sensing Models. It’s amazing that something so small can have such a huge impact!

In the coming weeks, I’ll share more details about our Small Sensing Models, what they are, what they do, and how they’ve come to change the way smartphone and PC makers are bringing AI innovation to their respective markets.

Make sure to stay tuned for more thoughts, news, and updates from me in the coming weeks and months!.

.