Sanjeevini.me - Healthcare App - Open Source - Audio AI

Started building an Healthcare app with Audio based AI.
Convert the conversation between Doctor and Patient and generate Reports.

Idea is to Integrate with Ayushman Bharat Digital Mission/ Aadhar Health ID.
Build via Open Source, to build features collaboratively.

web : https://sanjeevini.me
source - GitHub - slabstech/sanjeevini_frontend: UX for Sanjeevini.me - AI Health App

2 Likes

This is very cool. We have this healthcare enablement app on the playstore. It’s called Ensocure. I wonder if we could work together.

That’s amazing! As a non-tech founder, we’re struggling to find the right CTO co-founder to help us build a Food Navigator. Do you have any ideas or suggestions on how to approach the tech side of things? Also, how much did your app cost to develop?

I am the sole-developer building the prototype to production app.
Also i use Code assistants to integrate work and refactor my code.

I have been attending hackathon’s and talking to the dev’s to collaborate.
That would one of the ways to find collaboration.

For the prototype i am using free resources.
Frontend - React hosted on @github Pages Backend - Django hosted on @huggingface spaces, AI Inference - MistralAI and Nvidia NIM Free API Tier .

1 Like

Yes, We can collaborate for the project, I would love to get feedback for the base feature.

Idea is to build like GitLab/ OpenCore model, its a work in progress and still many things to do.
I am building it full-time now, a version should be ready by Mid-december 2024.

That’s a solid approach, @gaganyatri . Hackathons are indeed a great way to connect with passionate developers. Since you’re leveraging free resources for the prototype, that’s incredibly resourceful. Can you share more details on how you’re managing the integration of these tools?

Also, what would you suggest for a non-tech founder like me to get started with something similar, even if it’s just understanding the process better?

is an incomplete version of how my project is integrated.

I will update it soon, with some changes i have introduced.
You can use ChatGpt to expand the document and provide a roadmap to understand.

I also use Perplexity.ai a lot for adding new features and solving problems.

My dev setup is
VsCode + Continue Dev Plugin + Mistral API .

Discord for the project, though its not actively maintained.

Also my X account at x.com usually has some details.

Its a daily grind, try to experiment with small changes.
I dont copy anything more than 10-20 lines, it becomes impossible later to debug.

1 Like

Thank you

version 0.1 - Pitch document.

Collected all ideas and progress made over the last few months,
Will keep updating each month.

Drive Link

PDF below .
sanjeevini_pitch.docx.pdf (65.3 KB)

Thanks for the update

Benchmark for Medical Agents -

New healthcare support solutions using GenAI methods, could use this to validate their systems

MedAgentBench: A Realistic Virtual EHR Environment to Benchmark Medical LLM Agents
arxiv - [2501.14654] MedAgentBench: A Realistic Virtual EHR Environment to Benchmark Medical LLM Agents
source - GitHub - stanfordmlgroup/MedAgentBench: MedAgentBench: Dataset for Benchmarking LLMs as Agents in Medical Applications

For Sanjeevini.me - we have decide to Use Cloud hosting for demo and version.
Since the compute capacity of Jetson Nano was not sufficient for real-time traffic. Jetson Xavier NX is another option, though expensive. Without incurring additional development cost for on -device setups,
We will target large audience with Cloud API integration
Once smaller models are available, we will plugin our optimised code to deliver on-device solutions.

Thanks for the Update

We have contact !!!

After months of wrangling, finally got ASR(Automatic Speech Recognition) for kannada working .

Now we have - End to End Speech for Kannada.

No more walled Gardens !!!

Build your own Alexa/Siri/Google voice commands systems and speak with native languages.

All systems running locally with GPU and no continuous Internet required.

Business case :

mistral_large_kannada

1 Like

This is awesome @gaganyatri

Medical Transcription has to be accurate and verifiable by 3rd party audio,

We have now made the underlying technology Open Source and accessible to everyone.

Sanjeevini.me will be built on top of this stack.

Dhwani - Voice AI - Android app now available in Early access program.

DM, me your email-id connected to Google Play store to get access to the App.



Source -
Android - GitHub - slabstech/dhwani-android

Backend - GitHub - slabstech/dhwani-server