AI Database Migration Engine
A full Streamlit application I built by hand, with embedded AI endpoints — it takes on-prem SQL Server and auto-converts it into a modern cloud data architecture (Snowflake + dbt), wired into GitHub / Azure DevOps for CI/CD.
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
A complete application, built by hand: a Streamlit front end with embedded AI (Claude / GPT-4) API endpoints. Point it at on-prem SQL Server objects and it auto-converts them into the target cloud architecture — Snowflake SQL and dbt models, with the structure and tests a human engineer would write — then pushes the output straight into GitHub / Azure DevOps for CI/CD.
Results
- 100+ objects converted, tested at 85%+ accuracy
- Per-object migration went from weeks to hours
- End-to-end in one tool: conversion, version control, and CI/CD