No Signal, No Problem: Building a Vision-Assistive AI Application

Presented by Google Developer Group on Campus - Asia Pacific University Chapter

📅 Event Details
Date: Friday, 2 October 2026
Time: 4:30 PM – 7:00 PM (MYT)
Venue: TBC (APU Campus) + Online via Bevy
Format: Hybrid (In-person & Online via Bevy)
Duration: 2 Hours 30 Minutes
Department: AI Department
Workshop Lead: Samantha Tee Ker Xin

🎯 About This Workshop
Ever wondered what it would take to build an AI app that actually works with no internet, no API keys, no monthly bill? In this workshop, you'll build a fully offline mobile assistant designed for visually impaired users — one that uses your device's camera to describe the world out loud, and a conversational AI agent to handle tasks like reading calendars or sending messages. Everything runs on the phone. No cloud, no server, no catch.

👥 Who Should Attend?

• Mobile developers

• ML enthusiasts

• Accessibility advocates

• Students welcome

Prerequisites:

• Basic knowledge of Android development (Kotlin/Jetpack Compose)

• Flutter and Android Studio installed (emulator or physical device)

• Run flutter doctor before the session (should show green)

• Gemma model downloaded ahead of time (~1.5 GB) — a script in the starter repo handles this; please run it on home WiFi the night before

📚 What You'll Learn
✅ Edge AI fundamentals: cloud vs. on-device inference, and when each makes sense
✅ On-device vision ML: real-time object detection & text recognition using MediaPipe Tasks Vision
✅ Running an LLM on a phone: integrating Gemma into Flutter for fully offline reasoning
✅ AI agent + tool use: giving an LLM the ability to take real actions offline
✅ Accessibility-first design: building for voice and touch — and why it makes every app better

🛠️ Tools & Platforms
MediaPipe · Flutter · Gemma Agent

💡 Why This Workshop?
Most AI workshops end at the API call — send a request, display a response. This one goes further: it works where the internet doesn't, it keeps sensitive camera data on-device instead of a server, it's a rare skillset for MLOps/embedded AI careers, and it's hackathon-proof — no API keys, no inference bill, no demo that dies when the WiFi does.

What You'll Build

• Visual Tracker – point the camera at a room and the app describes what it sees, out loud, in real time

• AI Agent – a voice and text-driven assistant that answers questions, reads your calendar, and sends messages, fully offline

💬 Join the WhatsApp Group!
https://chat.whatsapp.com/JtqDzVBy1p64RSyzrO5duS

📧 Any Questions?
Connect with us through our social media channels and send us a message!