Blog

Notes from building AI agents, RAG and automation that run in production: what worked, what broke, and the trade-offs behind each decision.

First posts coming soon

In the drafts folder

  • Drafting

    Adding an LLM-as-Judge step to an extraction pipeline

    Why a second model checks the first one's output, and where a human still needs to review.

  • Drafting

    RAG over PDFs with tables that run across pages

    Chunking so that table rows keep their headers, and what that does to answer quality.

  • Planned

    Local or hosted LLMs for an internal RAG tool

    Thinking through hosting, cost and performance before a pilot goes live.

  • Planned

    From RPA bots to AI agents

    What carries over from years of process automation, and what changes when the logic lives in a model.