Polars
Rust-built DataFrame — Ritchie Vink (Xomnia), 2020. Faster than pandas (5–30×), uses Apache Arrow natively and lazy execution.
Two APIs: eager (immediate evaluation, pandas-like) and lazy (build a query graph, optimise and run with collect()). Multi-threaded by default, SIMD-accelerated. Scales on a single node to ~hundreds of GB; for larger data → Polars Cloud (managed distributed).
Competitors: pandas (established), DuckDB (the SQL route), Dask (distributed pandas), Modin. Polars wins ground quickly where pandas users hit memory limits.