Saarathi AI

India has hundreds of welfare schemes, yet millions miss out — SAARTHI bridges this gap by turning awareness into real access, guiding citizens from discovery to application with AI and CSC support.

Comment

Inspiration

We noticed something simple but serious — India has hundreds of government welfare schemes, but many people around us had no idea which ones they could actually use.

Even when they knew about a scheme, they were confused about eligibility, documents, or where to even start. In rural areas, this problem is even bigger because of language barriers and limited digital access.

We realized the issue isn’t the lack of schemes — it’s the gap between information and actual access.

That’s what pushed us to build SAARTHI.

What it does

SAARTHI is an AI-based system that helps people figure out which government schemes they can apply for, just by describing their situation in simple language.

For example:

“I am a farmer from Karnataka with low income”

Based on this, SAARTHI:

  • Understands the user’s background
  • Finds relevant schemes
  • Explains why they are eligible
  • Shows required documents
  • Guides them on how to apply

If the user is not fully eligible, it also suggests what’s missing.

The goal is not just to show schemes, but to actually help people move forward.

Accessibility focus

We tried to design SAARTHI keeping real users in mind:

  • Web interface (simple and clean)
  • Voice support (planned)
  • IVR system so people can use it without internet
  • Option to connect with a CSC operator for help

So instead of just information, the user gets a full path: from understanding → to action

How we built it

We built SAARTHI as a simple but structured system:

  • Frontend using Next.js (TypeScript + Tailwind)
  • Backend using FastAPI (Python)

The main logic works like this:

  • Take user input in natural language
  • Convert it into a structured profile
  • Match it with schemes from our dataset (schemes.json)
  • Return results with explanation and next steps

We also added APIs for handling assistance requests and operator workflows.

The dataset-driven approach helps us scale easily by just adding more schemes.

Challenges we ran into

One of the biggest challenges was handling user input.

People don’t speak in structured formats, so converting free text into usable data took effort.

Another challenge was explaining eligibility clearly. It’s easy to match schemes, but harder to explain why someone qualifies.

We also had to think beyond just the app — how this would work for someone without internet, which led us to the IVR idea.

What we’re proud of

  • We built a working prototype that actually completes the full flow
  • The system doesn’t just recommend — it explains
  • We included a path for real-world help (CSC operators)
  • The design is simple enough for non-technical users
  • The system can grow easily by expanding the dataset

What we learned

We learned that solving real problems is very different from building demo features.

Clarity matters more than complexity.

Also, technology alone isn’t enough — combining it with human support makes the solution more practical.

What’s next

We want to make SAARTHI more usable in real conditions:

  • Add support for multiple languages
  • Improve voice interaction
  • Expand the scheme database
  • Make eligibility explanations clearer
  • Add better tools for operators

Vision

We want SAARTHI to become something simple but powerful:

A tool where any person can understand and access the benefits they deserve without confusion or dependency.

Even if they don’t have internet.

Built With

Share this project:

Updates