In my day-to-day work with technology, I have found there is No Silver Bullet [ucsf.edu] to the essential difficulties of software development. Essential best practices are situational and defined by context.
This series describes how I made AI Search “Whatever data you want to see…” for Find Me in Chicago. Building Text-to-SoQL was a journey in pragmatic AI especially around the CHAD challenges: managing token Cost through minimal architectures suppressing Hallucination with strict schema and dialect grounding converting human Ambiguity into precise intent taming hardware non-Determinism […]
This series describes how I made AI Search “Whatever the data you want to see…” for Find Me in Chicago. I started my Text-to-SoQL journey with a naive misconception that cranking down these settings would make my queries mostly deterministic: temperature: 0 – sharpen the LLM to always choose the highest probability token topK: 1 […]
This series describes how I made AI Search “Whatever data you want to see…” for Find Me in Chicago. If a user asks to see the “worst crashes,” they are using a word that means different things to different people. Does “worst” mean the crashes with the most fatalities, the most injuries, or the ones […]
This series describes how I made AI Search “Whatever data you want to see…” for Find Me in Chicago. Human language is built on ambiguity. Ambiguity is not a bug to be fixed in natural language; it is an essential feature of human communication. As Ludwig Wittgenstein observed in his Philosophical Investigations (S43), “the meaning […]
This series describes how I made AI Search “Whatever data you want to see…” for Find Me in Chicago. LLMs are trained on billions of lines of SQL. SoQL looks like SQL but differs in critical ways: No `CASE`/`WHEN` keywords, although there SoQL does support a case() function No subqueries No `JOIN` supported for the […]
This series describes how I made AI Search “Whatever data you want to see…” for Find Me in Chicago. Everyone knows LLMs can confidently hallucinate and then often refuse to admit the hallucination unless prompted with clear evidence. So it is important to use schema grounding to give AI any chance to succeed at Text-to-SoQL. […]
This series describes how I made AI Search “Whatever data you want to see…” for Find Me in Chicago If you try to run “zero-shot” Text-to-Sql against a raw enterprise data warehouse with hundreds of tables and complex foreign key chains, off-the-shelf models fail spectacularly. Feeding hundreds of tables into a prompt pollutes the attention […]