Worked on the opensearch-project/sql repository to enhance query optimization and planner reliability, focusing on restoring dedup pushdown functionality when a WHERE clause is present. Addressed a complex bug by ensuring user predicates are preserved and correctly processed in the execution plan, using Java and SQL for implementation. Improved the deterministic behavior of the HEP optimizer by introducing per-rule instances and comprehensive regression tests, which stabilized rule firing across builds. Expanded end-to-end test coverage and explain plan validation, increasing confidence in production deployments. These efforts reduced scanned data and latency for dedup-heavy queries, resulting in more predictable and efficient query planning.