Ultrahuman, an Indian fitness startup, has partnered MongoDB, which provides an open-source NoSQL database management program, to collect and process real-time data, generated via its wearable devices, and provide health-related insights to end users.
Founded in 2020 by Vatsal Singhal and Mohit Kumar, Ultrahuman is a metabolic fitness company selling wearables, including Ultrahuman M1 and the Ultrahuman Ring.
While the ring has sensors to track temperature, heart rate and movement, M1 is embedded with sensors that collect glucose biomarkers via a CGM (Continuous Glucose Monitoring) device to help measure the impact of food, sleep, activity, and stress on the human body.
Singhal, told iTnews Asia, that both the M1 and the ring connect to the Ultrahuman app which feeds the data to backend servers where the data is analysed to provide real time health-related insights to end users.
“We push the collected data to servers located in the geography nearest to the user. The data is then cleaned and processed, after which health-related insights are shown on the app,” Singhal said.
Prior to the tie-up with MongoDB, Ultrahuman’s team of developers used to manage and update the data “manually” across each geography. Now the database automatically inserts millions of data points into the dataset in a second, allowing developers to “spend more time” improving products,” he added.
Ultrahuman is using MongoDB Atlas hosted on the AWS Cloud to process the “large volume” of data generated and provide real-time insights with low latency to health device users he said.
Singhal added that with MongoDB Atlas’s global writes capability, the “time” for data ingestion (collection of raw data from different source systems and loading it into the target system) has been reduced from over two seconds initially to less than 50 milliseconds.
“The data load has also become horizontally scalable.”
Simultaneously the time required for a user to visualise the data on the app has gone down from over one second to less than 100 milliseconds, Singhal said.
Scale with a single setup
The platform has over 150 million glucose data points, which the company claims to be one of the largest glucose datasets globally for healthy people, and it provides personalised AI-driven nudges to its users for measuring their performance via its app in relation to the stored data.
These glucose data points are “user data” generated from the wearables that the company is shipping.
Ultrahuman M1 is available in India and UAE and will soon be available in the United Kingdom.
Singhal noted that since the M1 uses a CGM, which is a medically-regulated device, there have been some temporary bottlenecks in rolling out the device in more countries.
The ring, on the other hand, is a hardware device and while it is available in India, Ultrahuman has plans to ship it worldwide and it already has some users in the US and UAE, Singhal said.
With a globally distributed user base there was a need to have separate setups in different geographies, he added.
With medical data sitting on its platform, Ultrahuman has felt the need to store the data locally and comply with the data privacy laws of respective countries or regions, Singhal said.
MongoDB’s Manager, Solutions Architect Corp APAC, Himanshumali, said the “sharded cluster” capabilities of MongoDB help Ultrahuman “eliminate the need” to create multiple setups across geographies.
“They can cater to the global audience through a single setup while ensuring that data is located locally in the region, as the database manages all by itself,” he said.
“The algorithm is inbuilt into the sharding methodology, so no application code has to be written, no logic has to be added from the application side, and the database takes care of it automatically,” he added.
The Ultrahuman app streams data in real-time to the cloud and it is “time-sensitive” data that is generated at regular intervals of either a few seconds or minutes.
MongoDB’s database stores this time-series data along with the “bucketing pattern”.
"While dealing with time-series data, real-time analytics, or the IoT-generated data, the time-series collection in MongoDB makes it easy," Himanshumali said.
This helps in discovering historical trends, providing forecasts for the future, and optimising storage, he added.
Also, with traditional databases, there is a need to build an operational setup and then “move data” to a different system for analytical operations, Singhal noted.
“This slows down the database and this is where MongoDB Atlas’s analytics node feature helps in workload isolation and provides real-time responses to users,” said Singhal.
Along with network isolation, the database configures role-based access, and encryption using transport layer security (TLS) over the network traffic – all these security features ensure a secure data environment, Himanshumali added.
MongoDB takes care of all the pain points and allows us to expand without worry, and that too, with the click of a button. “It helps us reduce costs, not just from a product perspective, but also from an engineering point of view”, Singhal said.
