Core Data Dashboard — Self-Service EDA

Every question about orders, patients, studies and files by facility, modality or time meant another hand-written multi-table SQL join.

A Streamlit app wrapping a seven-table join in one reusable function, with date ranges, facility-group filters, multiselects and regex search. The data engine was refactored from pandas to Polars; session caching means the expensive join runs once per session. Deployed on a GCP VM behind nginx with authentication.

• Order, patient, study and file breakdowns in seconds — usable by non-engineers.
• Bounded-memory ingestion via chunked reads; deployed and access-controlled.

• Directed that recurring analyst requests become a self-service product rather than ongoing engineering interrupts.