Wani-gra

AI For Meditation: How Headspace Leverages AI and ML Systems

Device understanding and synthetic intelligence is reworking the health care marketplace as we know it. Specially in the final calendar year and half, getting trapped inside of one’s residence has led to the manifold enhance in the number of people searching for out for digital wellness products and services. As a end result, the wellness business is mentioned to boost even more above the following 4 years. A development of $1299.84 billion is predicted concerning past 12 months and 2024, at a CAGR of 6.37 p.c through the forecast period of time. 

With a escalating field in sight, California-headquartered Headspace tapped into synthetic intelligence to increase overall health and wellness across the world. Established by previous Buddhist monk Andy Puddicombe, and Richard Pierson in 2010, Headspace tries to educate meditation and mindfulness at scale. The enterprise delivers guided meditation, mindfulness, snooze, workout, and focus content material by its world wide web, Android and iOS applications. 

Sign up for our future AI Conference>>

In its recent weblog write-up, penned by Yu Chen, Senior Software Engineer, and co-authored by Koyuki Nakamori, Senior Engineering Manager, Headspace reveals how it uses genuine-time device learning to continue to be at the top of its video game. 

Leveraging Information in Real-time 

Information is typically ingested, reworked into the sought after structure, persisted and then manufactured to sit idle right up until made use of by machine discovering engineers and analytics groups. Nevertheless, in purchase to make genuine-time conclusions, person details has to be leveraged promptly. To guarantee this, the Headspace crew has considerably shortened the end-to-conclude feedback loop. Consumer steps are analysed inside of seconds or minutes to produce appropriate, personalised, and context-precise tips. 

Headspace’s machine studying method incorporates capabilities that update throughout the day and even for the duration of sessions attended by every consumer. These attributes mainly immediate to: 

  • On-going session bounce prices for sleep content 
  • Semantic embeddings for user lookup terms. Indicating, if a user queries for ‘Preparing for exam’, the product will assign aim-themed meditations 
  • Biometric information these as step count and pulse of individual buyers help the model provide personalised workout content material

Tech Stack

The machine learning team at Headspace has designed a solution to cater to the personalised need of their buyers by breaking down the composition into publishing, receiver, orchestration, and serving levels. It leverages the pursuing technologies: 

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

Headspace works by using light-weight Lambda capabilities to pack and unpack details in appropriate formats and invoke Sagemaker endpoints to complete put up-processing and persistence. The architecture overview of Headspace is as revealed underneath: 

Source: Headspace 

The engineers make clear that functions created by consumers on the Headspace application are forwarded to the company’s Kinesis streams in buy to be processed by Spark Structured Streaming. The app then fetches predictions by making RESTful HTTP requests on its backend products and services. It also transfers user IDs and feature flags to point out the equipment finding out tips that need to have to be sent back again. 

Headspace re-trains its models by leveraging AWS deployment designs and updating the Sagemaker model. 

Blue Inexperienced Architecture

To steer clear of any disruptions throughout updates, Headspace has constructed a blue-green deployment model. That is, it maintains two parallel infrastructures or copies of feature suppliers. In addition, it designated one production ecosystem to route requests for characteristics and predictions towards it by means of Headspace’s Lambda. 

See Also


Supply: Headspace

Each and every time Headspace has to update its model, it takes advantage of a script to update the complementary infrastructure (as denoted by the blue atmosphere in the illustration above) with the newest attributes. When the update is carried out, the group switches the Lambda to stage to the up to date (blue) surroundings. The team thus keeps repeating the course of action just about every time it has to update the product. 

Thus, by enabling true-time inference, wellness company Headspace is equipped to significantly lower the end-to-end opinions loop between the person coming into the motion and the application providing personalised strategies. 

To know about how Indian startups are revolutionising the healthcare market working with synthetic intelligence and machine discovering, simply click listed here. 

Synthetic intelligence, device studying, meditation app, wellness application, Headspace, equipment finding out designs, healthcare and wellness industry


Be a part of Our Discord Server. Be portion of an engaging on line community. Be a part of In this article.


Subscribe to our E-newsletter

Get the newest updates and applicable delivers by sharing your e-mail.

Debolina Biswas

Debolina Biswas

Just after diving deep into the Indian startup ecosystem, Debolina is now a Technological know-how Journalist. When not creating, she is identified reading or participating in with paint brushes and palette knives. She can be arrived at at [email protected]