Cognition launched Devin for Databricks Migrations after one energy firm moved 142 objects in eight days.
Cognition contrasts that with work that once tied up an engineering pod for multiple quarters, Cognition says.
The pitch is time back for building.
Devin converts legacy code, tests it in the customer workspace, fixes failures and reconciles each source against its Databricks copy before cutover, Cognition says. Ali Ghodsi, co-founder and CEO of Databricks, said context trapped in legacy systems limits AI, adding that Devin and Unity Gateway move that context under Unity Catalog governance. Named inputs include Teradata, Oracle, SQL Server, Informatica, Redshift and Snowflake, with outputs in SQL, PySpark and Spark Declarative Pipelines.
Zeb cut a planned four-week NetSuite migration to two and a half weeks and built 20 tables for an investment advisory firm, Zeb CEO Mal Vivek said. Checks cover schema, row counts, aggregates and row-by-row diff before cutover, Cognition says.
Devin runs hundreds of tasks in parallel on a migration, Cognition says. Devin Cloud docs say teams encode migration instructions once, then agents fan out and open PRs for review.
Devin for Databricks Migrations is available on Databricks Marketplace, with purchases counting toward Databricks Universal Commit, Cognition says. Cognition did not disclose error rates, rework after cutover or cost compared with human-led migration.
Repeat reconciliation data would change the read. Will eight-day results hold across larger Teradata, Oracle and Snowflake estates?





