Experience with event-driven systems, ETL/CDC, PostgreSQL/pgvector, serverless architectures, Kubernetes, Vertex AI, Microsoft Foundry, or additional languages such as C# and Go would be beneficial, but none is individually required. Candidates should understand how to determine whether an AI capability is actually functioning correctly, including offline evaluation, asynchronous evaluation of deployed interactions, runtime validation, and the use of telemetry to investigate failures and measure quality.