AI For Meditation: How Headspace Leverages AI and ML Technologies

Machine understanding and synthetic intelligence is reworking the healthcare market as we know it. Especially in the very last yr and 50 percent, remaining stuck inside of one’s house has led to the manifold enhance in the number of men and women looking for out for virtual wellness expert services. As a consequence, the wellness industry is reported to maximize additional about the future 4 many years. A advancement of $1299.84 billion is anticipated in between last yr and 2024, at a CAGR of 6.37 per cent all through the forecast interval. 

With a escalating industry in sight, California-headquartered Headspace tapped into artificial intelligence to enhance health and fitness and wellness across the world. Launched by former Buddhist monk Andy Puddicombe, and Richard Pierson in 2010, Headspace tries to instruct meditation and mindfulness at scale. The company offers guided meditation, mindfulness, snooze, exercising, and target content by its net, Android and iOS applications. 

Sign-up for our upcoming AI Conference>>

In its new website put up, created by Yu Chen, Senior Software program Engineer, and co-authored by Koyuki Nakamori, Senior Engineering Supervisor, Headspace reveals how it takes advantage of real-time machine mastering to stay at the major of its recreation. 

Leveraging Facts in Serious-time 

Knowledge is typically ingested, remodeled into the ideal structure, persisted and then made to sit idle right until utilised by machine understanding engineers and analytics groups. Even so, in buy to make true-time selections, user information has to be leveraged immediately. To ensure this, the Headspace workforce has considerably shortened the close-to-close feed-back loop. Consumer steps are analysed inside of seconds or minutes to crank out suitable, personalised, and context-certain tips. 

Headspace’s equipment finding out mode incorporates options that update during the day and even for the duration of sessions attended by just about every person. These functions primarily immediate to: 

  • On-heading session bounce rates for slumber content 
  • Semantic embeddings for person lookup phrases. That means, if a person searches for ‘Preparing for exam’, the design will assign target-themed meditations 
  • Biometric facts these kinds of as phase rely and pulse of individual customers assistance the product supply personalised physical exercise articles

Tech Stack

The equipment studying staff at Headspace has made a remedy to cater to the personalised want of their people by breaking down the structure into publishing, receiver, orchestration, and serving layers. It leverages the pursuing technologies: 

  • Apache Spark Structured Streaming on Databricks 
  • AWS SQS 
  • Lamba 
  • Sagemaker 

Headspace takes advantage of light-weight Lambda features to pack and unpack info in ideal formats and invoke Sagemaker endpoints to perform put up-processing and persistence. The architecture overview of Headspace is as revealed underneath: 

Resource: Headspace 

The engineers make clear that activities created by consumers on the Headspace app are forwarded to the company’s Kinesis streams in get to be processed by Spark Structured Streaming. The app then fetches predictions by producing RESTful HTTP requests on its backend solutions. It also transfers consumer IDs and function flags to point out the machine finding out recommendations that need to have to be despatched again. 

Headspace re-trains its products by leveraging AWS deployment patterns and updating the Sagemaker design. 

Blue Inexperienced Architecture

To prevent any disruptions throughout updates, Headspace has developed a blue-inexperienced deployment design. That is, it maintains two parallel infrastructures or copies of attribute suppliers. In addition, it selected just one production setting to route requests for features and predictions to it by using Headspace’s Lambda. 

See Also


Source: Headspace

Each individual time Headspace has to update its model, it works by using a script to update the complementary infrastructure (as denoted by the blue setting in the illustration earlier mentioned) with the latest characteristics. Once the update is performed, the workforce switches the Lambda to position to the updated (blue) environment. The staff thus keeps repeating the system every time it has to update the product. 

Consequently, by enabling authentic-time inference, wellness corporation Headspace is ready to dramatically minimize the conclude-to-stop feed-back loop among the user getting into the motion and the application delivering personalised recommendations. 

To know about how Indian startups are revolutionising the health care field working with artificial intelligence and machine learning, simply click in this article. 

Synthetic intelligence, device understanding, meditation app, wellness app, Headspace, machine understanding styles, health care and wellness industry


Be a part of Our Discord Server. Be component of an participating on the internet community. Sign up for Below.


Subscribe to our Newsletter

Get the most recent updates and related gives by sharing your e mail.

Debolina Biswas

Debolina Biswas

Just after diving deep into the Indian startup ecosystem, Debolina is now a Engineering Journalist. When not producing, she is observed looking at or actively playing with paint brushes and palette knives. She can be reached at [email protected]