Introduction: A Story of Two Worlds
Think about a world the place you’re fearful a couple of new spot in your pores and skin, otherwise you need to verify your danger for a stroke, however you hesitate.
You hesitate as a result of you understand the second you entered that data into an app, it’s not yours. Your most private well being knowledge is uploaded, analyzed, and monetized — a personal fear become a public knowledge level.
Now, think about a unique world. A world the place your well being data stays precisely the place it belongs: with you.
The Downside and My Resolution
I’ve constructed an app that brings this second world to life. It’s an AI-powered software that gives early detection evaluation for pores and skin most cancers and stroke danger, all from an area gadget. The magic is in its design: it really works fully offline.
You should use your cellphone’s digicam to research a pores and skin lesion or reply 10 inquiries to assess your stroke danger. All of this private and personal data is shipped to an area machine studying mannequin proper in your gadget.
That is extra than simply an app; it’s a brand new paradigm for private well being. It’s about personal well being checks, early detection, and your personal offline LLM.
Consider it as bringing the spirit of Baymax to life — a trusted ally that’s all the time there, no web required, and fully devoted to your well-being.
The magic occurs because of a very native setup, using Gemma 3N 4B in your gadget, a system I’ve constructed from the bottom up (sure, with Vibe).
Right here’s a fast demo of the app in motion:
https://www.youtube.com/watch?v=Mr7RlW_o7Co
On this demo, you’ll see a easy HTML/JavaScript/CSS app — which you’ll be able to simply redesign your self — speaking on to Ollama, which is put in domestically in your Android gadget or laptop computer. It then makes use of Gemma 3N 4B to offer a abstract of the machine studying evaluation.
That is all to show that superior, personal AI isn’t just doable — it’s right here.
The Massive Query: To Promote, or To Share?
As I developed this mission, I confronted a vital determination that many creators grapple with: How do you share a software like this with the world?
Do you attempt to monetize it by an app retailer, aiming for a startup, and even search enterprise capital? Or, do you share it freely, trusting within the energy of the group?
My selection was clear: I made a decision to share it.
Why Sharing Issues:
My purpose is to make these life-saving instruments as accessible as doable, whereas concurrently reinforcing the significance of private knowledge privateness. By open-sourcing the code, I hope to encourage others to construct on this basis.
You could find the complete supply code and my machine studying fashions right here:
My machine studying fashions on Kaggle:
ML_cnn-skin-cancer-classification
Supply code to my offline Ollama/Gemma 3N fashions: github.com/git-prototype-ai/kaggle
I’ve constructed this to show that superior, personal AI isn’t just doable — it’s right here. I hope this work conjures up others to construct upon it, as a result of collectively, we will put Your Well being, Your Privateness, and Your Peace of Thoughts again in your management.
I might love to listen to your ideas. Was open-sourcing the correct selection for a mission with this type of impression? What are the opposite paths for monetizing such an thought whereas staying true to its core values?
