AI Database Migration Engine
An autonomous LLM agent that refactors SQL Server stored procedures, views, and T-SQL into Snowflake SQL + dbt models — 100+ objects at 85%+ accuracy, cutting migration from weeks to hours.
Applied AI LLM Agents Snowflake dbt
The problem
Migrating a legacy SQL Server estate — hundreds of stored procedures, views, and T-SQL objects — into a modern Snowflake + dbt stack is normally a slow, manual, error-prone slog.
What I built
An autonomous LLM agent (Claude / GPT-4) that reads each SQL Server object and emits the equivalent Snowflake SQL and a dbt model — with the structure and tests a human engineer would write.
Results
- 100+ objects tested at 85%+ accuracy
- Per-object migration time went from weeks to hours
- Output treated as software: version-controlled, testable, reviewable