We explore wavelet trees as a space-efficient alternative for physical-to-sorted mapping in learned indexes, demonstrating how traditional data structures can be evolved to diverse applications and evaluate the space-performance tradeoff observed in systems.

A task decomposition approach to learned LSM tuning that breaks down the complex tuning problem into two sub-tasks - a cost model used that evaluates a given tuning and a tuner that suggests knob values based on the cost model.

My first conference talk at KubeCon about Elevating Cloud Native Education.

A week long data intensive workshop hosted by Goetz Graefe