Worked on the IBM/velox repository to enhance support for complex nested data types within columnar data processing systems. Developed new features in C++ that enabled the ColumnSelector to handle both MAP and ROW types for flatMap columns, broadening validation logic and improving expression handling for nested structures. Expanded benchmarking utilities by introducing fuzzMap support and a random map type generator, streamlining map generation for performance testing. Focused on robust software testing and fuzzing techniques to validate correctness when filtering with Row types and processing nested arrays, contributing to more reliable query execution and performance tuning for advanced data processing workflows.