Thor Schaeff works on developer experience for the Gemini API at Google DeepMind. His tutorials connect model APIs, application code and the infrastructure needed to turn a demonstration into a usable product.[3]
Start with Self-host Maps with Protomaps and Supabase Storage →
Links
WebsiteMember of the Technical Staff (DevX) · Google DeepMind
Career
Schaeff previously worked at Stripe, Supabase and ElevenLabs. His professional timeline describes building developer experience and developer relations functions at Supabase, followed by work on ElevenLabs’ audio developer platform. At Google DeepMind, his focus includes audio understanding, speech generation and the Live API. These roles connect practical application engineering with teaching developers how to use a platform.[2]
Contributions
Voice systems that keep listening
Schaeff coauthored Google’s introduction to Gemini 3.8 Live and 3.5 Transcribe. The Live models support voice agents that continue a conversation while external tools execute in the background. The announcement also describes visual context and incremental content updates, which let an application combine a live audio stream with information from other systems. His credited contribution is the developer announcement and its examples.[3]
Maps as a deployable file
With Brandon Liu, Schaeff wrote a guide to self-hosting maps with Protomaps and Supabase Storage. It starts with a selected geographic area, packages map tiles into a static PMTiles file, and renders that file with MapLibre. The tutorial then adds an Edge Function to restrict access. This makes the map’s storage and authorization behavior visible in ordinary web application code.[4]
Tools, writing & talks
Writing, research & talks
Start here
A complete path from extracting map tiles to rendering a browser map and controlling file access.
Start with the map guide to see his teaching style in a concrete application. Basic JavaScript and a Supabase project are useful; no machine learning background is needed.
Coauthor[4]A Google DeepMind developer-experience talk on building real-time voice agents with Gemini. Covers the Live API's native-audio WebSocket sessions, barge-in, tool calls and grounding, plus stateful image chaining and live translation demos.
Catalog workOriginal sourceIntroduces native voice interaction, background tool execution and transcription in Google’s audio API suite.
Coauthor[3]