Oysty India
Issue #03AI & Mobile2026

Private AI Assistant

Meeting notes that never leave your phone

A free, open-source Android app that records meetings, transcribes them with on-device Whisper, summarises them with a local Llama model and answers questions about your meeting history — without sending a single byte to the cloud.

KotlinAndroidWhisperLlama 2.5On-device MLMaterial YouMCP
CASE
STUDY

The challenge

What was broken

Meeting notes usually live inside cloud transcription services, which means sensitive calls, client conversations and internal stand-ups are uploaded to somebody else's servers. The goal was to keep every useful outcome — transcripts, summaries, action items and search — with none of the upload, no account and no internet connection required.

The solution

What we built

We built a fully offline Android companion around on-device models: Whisper for speech-to-text, a local Llama 2.5 model for summaries and chat, and a local index for semantic search across past recordings. Teams that want extra reach can plug in their own OpenAI-compatible provider or add MCP servers from the built-in directory.

Inside the issue

Key features

The parts of the product that do the heavy lifting every day.

  • On-device Whisper transcription (tiny, base or small model)
  • Local Llama 2.5 summaries with topics, action items and decisions
  • Natural-language search across every past meeting
  • Meeting Q&A chat grounded in your own transcripts
  • No accounts, no cloud, no ads and no data collection
  • Biometric app lock, dark theme and Material You design
  • Resumable, checksum-verified model downloads
  • Optional OpenAI-compatible providers and an MCP directory
From the build

Real screens

Screenshots from the shipped product — not mockups and not stock imagery.

Private AI Assistant home screen with a record button, timer and bottom navigation for recordings, AI chat, models and settings
Record a meeting — everything is processed on the device
AI Meeting Chat screen showing a question about key topics and an answer generated from the meeting transcript
Ask questions and get answers grounded in your own recordings
Recordings screen listing a saved meeting with its duration and file size
Meeting history with durations and sizes, searchable offline

The utility belt

Technology used

  • Kotlin
  • Android
  • Whisper
  • Llama 2.5
  • On-device ML
  • Material You
  • MCP

We choose boring, proven tools where they help and modern ones where they earn their place. Every stack decision is documented so your team can maintain it later.

The aftermath

What changed

  • Sensitive meeting audio never leaves the device
  • Transcription, summaries and chat all work offline
  • Free and open-source, with no tracking or ads
  • Published on Google Play and shipping regular updates
Kaboom!

Ready to start your project?

Tell us what you want to build. We will reply within one business day with honest advice and a clear next step — no pressure, no jargon.