The lakehouse architecture is designed to enable interactive analysis of uncurated and curated data stored in data lakes. By adopting a lakehouse architecture, companies can minimize storage costs and circumvent the need for some ELT/ETL pipelines. However, developing high-performance query engines that match or exceed the performance of data warehouses on raw data and structured data is challenging. In response to this challenge, Databricks developed Photon, a vectorized query engine written in C++ that executes queries written in SQL or the Spark Dataframe API. By adopting vectorized execution, Photon can support runtime adaptivity to discover and exploit the characteristics of micro-batch data, thereby improving query performance on raw data. Photon integrates with the Databricks Runtime at the operator level so that users can run their existing workloads without modification and queries can fall back to SparkSQL for operations that are not yet supported.