Local AI infrastructure

in progress

A local AI service layer for personal applications and assistants. I want one useful, private inference setup instead of a different pile of scripts for every idea.

  • Ollama
  • private inference
  • shared APIs

On the bench

  • Making local inference reusable across small personal tools
  • Learning where latency and model choice change the experience
  • Keeping the service boundaries loose enough to experiment

Grammar assistant

experiment

A Chrome extension that connects browser text fields to a local LLM for German grammar feedback, helping me learn from recurring mistakes. Unfinished drafts stay on my machine that works as a server; the extension uses the shared inference layer rather than managing its own models.

  • browser APIs
  • local AI
  • German grammar

On the bench

  • Finding feedback that is useful without becoming annoying
  • Keeping local processing visible and trustworthy
  • Remembering recurring mistakes without pretending the model is always right