Databricks research shows multi-step agents consistently outperform single-turn RAG when answers span databases and documents

Data teams building AI agents keep running into the same failure mode. Questions that require joining structured data with ...
GitHub

SIGMOD 2026 · Software-Engineering-Inspired Text-to-SQL

🧠 Software-engineering pipeline instead of monolithic prompt-only generation. 🧪 Checker-based SQL revision for syntax, execution, and result-level repair. 🎯 Execution-aware selection over multiple ...
GitHub

sql_to_build_pipeline.sql

-- Create table that will hold the remaining 990k records to be inserted into the landing table, simulating streamed-in data. CREATE or REPLACE TABLE INSURANCE.ML_PIPE.INCOMING_DATA_SOURCE ( -- * Run ...
IEEE

Automated Traffic Video Analysis with a Modular Computer Vision Pipeline and SQL-Based Analytics

Abstract: The increasing volume of traffic video data demands efficient methods for extracting actionable insights to improve road safety and traffic management. This research presents an automated ...