Aisylum

Banner image
Landing page, recent actions
Translate text
Document extracted text
not so unique chat interface

Inspiration

We live in a world where access to information can mean survival, yet privacy is often a privilege. With numerous major conflicts globally, this is even becoming more of a cause of concern. Refugees, asylum seekers and people in restricted regions also deserve access to the power AI tools can offer without exposing them.

What it does?

Aisylum is a privacy-first, offline-ready AI platform that gives users access to chat with AI, analyse documents, and learn languages without ever compromising their data. It’s a space where you can think, learn, and communicate freely - powered by Gemini Nano, protected by design.

How I built it

I built Aisylum as a Progressive Web App (PWA) - frontend heavy (if not only), lightweight, installable, and functional even offline.

Offline Storage: It uses local storage, the browser's database (IndexedDB) and on-device caching to keep data private. In-Browser AI: Using Gemini Nano and Chrome's built-in AI, I summarized and filtered content locally.

extra: Fallback to Gemini api handling requests - so others can explore the platform

Challenges I ran into

Honestly, the greatest one I faced was that my laptop isn't even able to run the models so I had to develop on mine and test with my mom's :)

Accomplishments that I'm proud of

I built a working AI platform that runs independently, protects users’ data, and proves that privacy and intelligence aren't mutually exclusive.
Funnily enough the name, Aisylum (Not the worst name I've formed tbh).
An elegant refactor that allows to use the same interface for fallback remote AI call.

What I learned

I learned a lot of how much of horizontal distribution you can achieve when the client-side can offload part of the strain form the server-side.

I learned that ethical technology isn’t just about encryption or APIs - it’s about empathy. Designing for people in vulnerable contexts forces you to rethink assumptions about access, trust, and control. I also discovered the power of local-first design - small decisions (like storing data on-device) make a huge impact on privacy.