Bobb KimYour AI Can Read Your Docs — But Not Your TablesTables are the blind spot in RAG pipelines. We rebuilt how Notion and Confluence tables reach your LLM — from Markdown to self-describing header-value pairs — and improved answer accuracy by 20.6%.Read moreYour Slack MCP Isn’t Broken — Your AI Just Can’t Read ItConnecting Slack to an AI agent is easy. Making that data actually useful to an LLM is the hard part. We improved retrieval accuracy by 27% with three changes: thread-level document splitting, noise filtering, and markup cleanup.Read more
