Minneapolis AI Engineer Meetup

What does it take to run AI on devices small enough to fit in your pocket? And when you need speech-to-text in production, should you reach for a cloud API or run models locally? This month, we're tackling two sides of on-device AI: shrinking LLMs down to mobile size, and choosing the right speech recognition approach for real-world constraints.

Colin Lee, Staff Mobile Engineer at webAI, takes on one of the most practical challenges in AI engineering: running LLMs and RAG entirely on a mobile phone. Colin brings experience from Mozilla, Meetup, Amazon, Flipgrid, and When I Work. He'll break down quantization, pruning, and distillation—the techniques that make on-device AI possible without cloud dependencies.

Then Jaim Zuber, Apple Platforms Engineer and Leader, surveys the current speech-to-text ecosystem—from cloud APIs like Deepgram, AssemblyAI, and Baseten to on-device models like Whisper, Parakeet, and Moonshine. He'll compare cloud and local approaches and dig into the constraints teams hit in production: reliability of long-running audio streams, privacy requirements, domain vocabulary, multiple languages, and accuracy. The session closes with practical guidance on how engineers can start building and testing ASR-based applications today.

Whether you're exploring edge AI to cut latency and protect privacy, or evaluating speech-to-text options for a production application, this session gives you practical knowledge you can apply immediately.

Agenda
Wednesday, April 8, 2026
🥨 4:00 PM – Doors open + snacks and networking
👋 4:30 PM – Welcome and intro remarks
🎤 4:40 PM – Colin Lee presents: Micro Machines Learning: LLMs and RAG Pocket-Sized
🎤 5:10 PM – Jaim Zuber presents: Speech-to-Text AI Models: From Cloud to On-Device — What Actually Works in Production
💬 5:40 PM – Networking and discussion
🚗 6:00 PM – Go home and build