Inspiration

Databricks DAIS 2026 Hackathon provided us with a platform to explore agentic AI build and deploy capabilities, and the 10K+ dataset of real care facilities from India provided in this event inspired us to explore our capability to develop an AI for good functionality that can be intuitive enough to be used by non-tech users at location in India.

What it does

The app helps you answer following question -

Can a care facility actually do what it claims?

  • The "CareTrust India" App evaluates facility claims for capabilities 7 key capability claims, such as ICU, NICU, maternity, emergency, oncology, trauma, or Dialysis.
  • For each facility and capability, it produces a clinical info trust score based on data availability for following attributes - capability claims, equipment evidence, procedure evidence and capacity data.
  • We also have social media trust score based on the data about official website, social media presence, last post, number of followers, etc.

Where are the highest-risk gaps in care, and how confident are we that those gaps are real?

  • App aggregates trust-weighted facility evidence across geography, and presents the evidence based score, identifying high-risk gaps in care by showing the lowest scored facilities in selected criterion

Where should a patient or coordinator actually go?

  • App allows the user to enter a location and a care need, such as “dialysis near Jaipur” or “emergency surgery near Patna,” and provides an evidence-attached shortlist of candidate facilities.

What needs to be fixed before this dataset can be trusted for planning?

  • The App profiles overview contradictions, suspicious claims, sparse fields

How we built it

Using Databricks Apps and Databricks Managed tables

Challenges we ran into

  1. Data Quality is the biggest challenge especially the geographical data was not accurate for 60% of records
  2. Ramping up skills to build and deploy Databricks Apps

Accomplishments that we're proud of

  1. Building a fully functional app to support evidence based trusted care by location
  2. Planned ways we can go above and beyond the item asked in the track by adding the quick search feature, providing trust scores separated by clinical info data and social media presence scores

What we learned

How to build Databricks Apps from scratch and iterate to keep improving them How to leverage Genie Code to expedite our development and testing efforts

What's next for CareTrust

  • We can convert this into a fully functional production grade app which can support CDC and real time updates to the data from care partners
  • Add more functionality for location based search by maps, and feedback capture capabilities in future versions

Built With

  • claude
  • database
  • databricks
  • geniecode
  • lakehouse
  • python
  • sqlwarehouse
Share this project:

Updates