AI in Database Modernization: Streamlining On-Prem to Babelfish Aurora PostgreSQL Migrations

Modernization of tradition on-site, in terms of scalability, performance, and data consistency, SQL Server systems present major difficulties. This work offers an integrated Artificial Intelligence (AI)-powered framework intended to automate the conversion of these databases to a cloud-native, Babelfishenabled Aurora PostgreSQL system. Deep learning (DL) transformer models (T5/BERT) at the core of the framework precisely translate complex SQL Server schema elements, including tables, columns, constraints, indexes, and stored procedures, into their PostgreSQL equivalents, so addressing problems including datatype mismatches and procedural logic conversion. Complementing this, a reinforcement learning module iteratively refines query execution plans by adjusting to Postgres' particular indexing and performance tactics, hence improving query optimization. To ensure data integrity and find potential errors such as missing foreign keys or ineffective indexing, an autoencoder-based anomaly detection system is also used. Although stored procedure translations provide more difficulties with an accuracy of 87 %, experimental results show that important schema components such as data types, indices, and constraints are maintained with above 95 % accuracy. All things considered, the suggested architecture not only reduces human involvement but also offers a scalable and strong answer for database modernization, hence improving operational efficiency and migration process performance.