Slack data wizard

Every company has data. Most don't have a data team. That gap is the market. In a small business, the people who own the data can't query it. The office manager holds the signup sheet. The clinic administrator holds scanned invoices. Neither writes SQL; neither has an engineer to ask. So the question never gets asked. Slack Data Wizard is that data team. Drop a CSV and it becomes a typed table, named in your words. Ask "how many signups per country?" and Gemma writes the SQL. Say "create a schema called sales" and Gemma writes the DDL. Say "build a medallion pipeline" and get bronze, silver and gold tables. Say "create a dashboard" and Gemma picks the chart and publishes a real Tableau workbook. Ask for "the top 10 countries by population" and Perplexity pulls the real figures from the internet, with citations — or OpenAI generates synthetic rows for testing. You can even ask by voice through an ElevenLabs agent, or reach the same tools from Claude and ChatGPT via MCP. Then drop in a scanned PDF — the case that matters most. Hospitals and billing offices still run on paper. Today a human retypes it. Gemma's vision reads the table straight off the page. And that changes who touches the data: the person who produces it loads the lakehouse directly — no analyst, no engineer, no untracked copies in between. For sensitive records, the fewer people in the chain, the smaller the leak surface. So the AMD GPU is not an implementation detail. A scanned patient record is exactly the data you may not paste into a closed third-party API — HIPAA, GDPR. An open Gemma on your own AMD Instinct GPU (ROCm) means the page never leaves the building. Take it away and in regulated industries you get no product. A model that writes CREATE TABLE can also write DROP TABLE — so every statement is classified before it runs, and destructive ones wait for a click. This deserves to win because it solves a real problem: it makes data handling faster, simpler and accessible to everyone.