Developed a Python API enhancement for the lancedb/lance repository, enabling zero-copy access to IVF centroids produced during index construction. Leveraging Rust, Python, Arrow, and PyO3, the work exposed internal centroid data directly to Python users, ensuring bit-level parity with the underlying search routing logic. This approach eliminated the need to re-run k-means clustering in Python for analytics, reducing both computational overhead and latency. By integrating with the existing FFI layer, the solution allowed richer in-database analytics without data duplication or recomputation, supporting more efficient workflows for users requiring exact centroid information within LanceDB’s analytics environment.