tinyML Talks on March 21, 2024 “Unlocking new frontiers in wearable technology with TinyML” by Kenneth Joel

We held our next tinyML Talks webcast. Kenneth Joel from Sensio Enterprises presented Unlocking new frontiers in wearable technology with TinyML on March 21, 2024.

Kenneth Joel from Sensio Enterprises explains their design philosophy for ultra low power embedded systems intended for wearables and how TinyML is crucial to ensure tight power constraints are met.

In this talk we explore the innovation behind their latest product the Orbyt Smart Ring Orbytring.com which measures multiple physiological parameters using a combination of multiple biosensors and runs on a battery that’s a fraction of the size you’d typically find on a smartwatch.

From creating his own sensor based shoes to creating AI-enabled software and Algorithms for Sensio, Kenneth shows excellence in his craft. He brings a fresh perspective to Sensio and the MedTech industry, thanks to his wide-ranging interests from music and fitness to technology. He is passionate about Sensors and AI. Kenneth has been an inventor from a young age, and is keen to create scalable solutions to large problems faced in the industry. At Sensio, Kenneth holds charge of all things Software, Process and HR, also contributing to Business Development. He holds a B.Tech Degree in Electrical and Electronics Engineering from PES University.


Download presentation slides:
Kenneth Joel

Watch on YouTube:
Kenneth Joel

Feel free to ask your questions on this thread and keep the conversation going!


Q: Is there an interface or API that the orbyt ring offers so that external developers can access anonymized data to process and add value to?
A: Yes we do plan to make APIs available so others can build apps and integrations on the Orbyt Platform. Please connect with me on email and share your detailed requirements. You can reach me at kenneth@sensio-ai.in

Q: Greetings! You talked about the lack of available sensor data being a bottleneck in TinyML development. What do you think causes this bottleneck aka why there is this lack of data when there are basically millions and millions of sensors in every possible device all around the world?
A: There is plenty of data available but there is a real dearth of high quality labelled datasets in the public domain. This is mainly because labelling of sensor data normally requires deep domain expertise or the availability of a simultaneous recording from a gold standard reference device as in the case of medical sensors. This is quite unlike the case of voice and video use cases which already have large public datasets where any trained human can fulfil the role of data annotation.

Q: 1. Why not a smart bracelet rather than a smart ring? Given that it’s just an additional smart device to make our lives better?
2. What’s the UVP/USP that Sensio has that other competitors do not?
3. What’s the life of this ring? As in, LTV of the product?
4. What are the additional medical param’s that will be introduced with this ring?
5. Is this product meant more for athletes? If not, will there be a specific model coming out for them?

A: 1. Answered in detail in the section titled “Why a smart ring”
2. World first Smart Ring to add an ECG sensor. Cloud architecture built around dynamic time bound consent to allow for flexible data sharing with loved ones and caregivers.
3. Device Life and Lifetime Value are two different things, please clarify.
4. HR, SpO2, Temp, ECG, Sleep Tracking, Activity Tracking
5. This product is for anyone who wants to improve and track their own health or the health of their loved ones.

Q: With ML algorithms embedded in the wearable devices, can we retrain the algorithm?
A: Yes we can retrain the algorithm and provide updates as required.

Q: Even though the ring is just coming to the market, do you have any visions or plans for the future where your company integrates with other healthcare systems/services etc. as a future approach to health monitoring/mgt?
A: Yes we do plan to integrate real world diagnostic tests into our offerings. These tests can then be used to reinforce and customise the insights offered by the ring. Sleep, Stress and Blood Pressure are of particular interest to us. Can’t reveal much more than this! Stay tuned to orbytring.com for more updates :slight_smile:

Q: Great presentation, Kenneth! Good luck on the launch of the product…How did you validate the sleep data processing as it has not been scientifically proven great with other devices? Which is the holy cow for validations?
A: Thank you! The medical gold standard for sleep monitoring is polysomnography (however even this gold standard has only 80% interscorer agreement) So I’d say the real gold standard for algorithm development and validation is a Polysomnography report scored by multiple technicians.

Q: What do you mean by annotation, is this labeling?
A: Yes

Q: Any chance for checking sugar levels? I assume pricking a finger is still being used
A: Maybe in the near future. Short-wave infrared (SWIR) spectroscopy is already looking promising in the wrist watch form factor. So maybe around 2027-28, with limited accuracy and need for frequent recalibration like existing cuffless BP solutions.

Q: Transformer models have been state-of-the-art architecture across multiple domains, are they popular when it comes to sensor data?
A: As far as I am aware, no commercial sensor based device uses transformer models.

Q: Access to the data is essential So third party developers aps can build upon the value of your smart ring provides. Also, how is cortizone measured by your ring?
A: Cortisol is not measured by the ring. However cortisol tests can be used to reinforce stress scores detected by the wearable (typically estimated using Heart Rate Variability and Resting Heart Rate)

Q: How do you prioritize which metrics are most valuable to users, and which insights should pull max power?
A: Every user defines their health goals during the onboarding process, the app experience is then customised according to their goals.

Q: Thanks for the presentation and insights - very interesting!
A: Thank you! Please join our waitlist at orbytring.com

Q: Thank you for the great presentation! GL for the product launch!
A: Thank you! Please join our waitlist at orbytring.com

Q: Sensors can be very sensitive and noisy in data collection, how is Sensio ensuring consistency in measurements across devices/sensors?
A: Yes, we invest much more in data collection, cleaning and qualification than we do in model development :slight_smile:

1 Like