❄ SNOWFLAKE CORTEX FEDERAL DEMO // SYNTHETIC DATA // USAF READINESS USE CASE
❄ Snowflake Cortex · Federal demo

Bring AI to governed mission data.
Don't move mission data to AI.

"Don't move mission data to AI. Bring AI to governed mission data."
WrenchLens is a Snowflake Cortex Federal demo. Every LLM call lands inside the customer's Snowflake account: the prompt, the grounding rows, and the response stay under the same RBAC, masking, and Query History audit as the source tables. Use case below is USAF aircraft readiness over a synthetic AF_READINESS schema.
❄ Snowflake Cortex live AI
SNOWFLAKE.CORTEX.COMPLETE, SUMMARIZE, and CLASSIFY_TEXT run inside the account. No external API calls, no third-party LLM proxies, no new ATO surface.
SELECT SNOWFLAKE.CORTEX.COMPLETE(model, prompt)
📊 Governed mission data in Snowflake
Aircraft, work orders, parts, and discrepancies live in AF_READINESS views. Cortex prompts are grounded in those views — one source of truth, no copy-out to a vector store.
V_FLEET_SNAPSHOT · V_OPEN_DISCREPANCIES · V_PARTS_RISK
🛡️ RBAC · Masking · Query History audit
Roles gate which tails and shops a user can see. Dynamic Masking redacts PII on the way out. Every Cortex call is a row in SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY — fully auditable.
ROLE · MASKING POLICY · QUERY_HISTORY
📈 Streamlit-in-Snowflake option
The same Cortex calls power a Streamlit-in-Snowflake app (streamlit/streamlit_app.py) — zero egress, native auth, deploys inside the customer's account.
streamlit-in-snowflake · native auth · zero egress
Compliance posture. Snowflake on AWS GovCloud carries FedRAMP HighDoD IL5 authorizations. Cortex LLM functions are on the FedRAMP High / IL5 roadmap — this demo is built so the same code runs on a Cortex-authorized region the day the agency turns it on. No claim is made that Cortex itself currently holds an IL5 ATO.