Tomás Senart, a principal engineer at Axiom, shares his holiday-driven experimentation with optimizing the performance and efficiency of Axiom's columnar database, which is designed to be cost-efficient and time-focused. Axiom allows for variable schema databases where events can have different fields and types, which enhances ease of data ingestion and integration. The architecture partitions data into blocks for storage and querying, using a custom Decomposition Storage Model (DSM) format that optimizes compression. Senart's recent work involved significant contributions to bitmap serialization, reducing costs and latency while maintaining performance. He experimented with dictionary encoding for _time columns, discovering that using dictionary encoding and frame of reference encoding can significantly reduce memory usage and improve performance. Although challenges remain, such as dealing with datasets with wide timestamp ranges, Senart remains optimistic about future improvements, driven by empirical validation and a passion for making systems faster.