NeuroPocket

• 

  Application UI

• 

  Application Greeting

Inspiration

Traditional AI health assistants and medical chatbots rely heavily on cloud infrastructure. This model presents two massive roadblocks: privacy concerns regarding sensitive medical inquiries and a reliance on internet connectivity, which fails in remote areas or emergency situations.

We wanted to change that. Our inspiration was to build a deeply intuitive, medical-companion app that puts the power of a modern Large Language Model directly into the user’s pocket—guaranteeing that personal data never leaves the device while maintaining accessibility anywhere on Earth.

What it does

NeuroPocket is a fully offline, private AI health assistant for Android. Upon launch, it opens up a fresh, distraction-free chat where users can securely inquire about health guidance.

Key Features:

• 🤖 On-device LLM — Runs Google's Gemma 1.1 2B model locally via MediaPipe
• 🔒 100% Private — No data ever leaves your device
• 💬 Multi-conversation — New chat on every app launch; full history in a sliding drawer
• 🏷️ Auto-titled chats — Conversations are automatically named from your first message
• 📂 Chat history drawer — Swipe or tap ☰ to browse and switch between past conversations
• ⚡ Optimised & smooth — Token streaming, auto-scroll, background inference, memory limits
• 🎨 Premium dark UI — Glassmorphism top bar, animated pulse indicator, gradient send button

🏷️ Built With

• android
• kotlin
• jetpack-compose
• mediapipe
• gemma-2b
• coroutines
• sharedpreferences

Built With

• android
• coroutines
• gemma-2b
• jetpack-compose
• kotlin
• mediapipe