A logistics company ingests 15,000 vehicle telemetry events per second that must be filtered, enriched, cataloged for SQL, and available to analysts within minutes. Engineers want no servers to manage and pay only for queries run. Which serverless data processing and analysis pipeline meets near-real-time and ad hoc needs?

The correct pipeline succeeds because Kinesis Data Streams scales to 15,000 events per second with low-latency ingestion, Lambda provides serverless event-driven filtering and enrichment without capacity planning, continuous delivery to S3 creates durable low-cost storage within minutes, Glue Data Catalog makes new partitions discoverable for SQL, and Athena provides true pay-per-query ad hoc analysis with no infrastructure to manage. The Firehose to scheduled Glue jobs approach fails on freshness because hourly batch transformation and catalog updates cannot make records queryable within minutes even though ingestion, storage, and Athena remain serverless and economical. The S3 plus Glue plus provisioned Redshift variant fails on operating model because a provisioned warehouse requires node sizing, maintenance, and always-on billing, which violates managing no servers and paying only for queries despite strong SQL capability. The Lambda to DynamoDB with PartiQL variant fails on analytics pattern because DynamoDB is an OLTP key-value store optimized for key lookups, not for cataloged full-dataset scans and aggregations across billions of telemetry events. As an exam caveat, treat words like scheduled, hourly, nightly, or provisioned cluster as decisive disqualifiers when the stem explicitly demands within minutes, serverless, and pay-per-query. As a practical check, measure event-time to Athena-queryable latency with expected partition volume and confirm Glue partitioning plus Athena workgroup limits sustain both freshness and query cost at peak throughput.