Benchmarks

No unverified performance claims. Reproducible measurements will be published only with the evidence needed to repeat them.

Methodology in preparation

The previous illustrative counters have been removed. Until the benchmark suite and raw results are public, this page intentionally shows no speed, memory or competitor figures.

Representative datasets

Small, medium and large tabular inputs with mixed numeric, categorical, text and datetime columns.

Workflow-level timing

Profile, assess, fix, prepare, validate and EDA measured independently and end to end.

Memory and scaling

Peak memory, row-count scaling and the effect of optional input formats recorded separately.

Reproducible evidence

Every published number will include code, dataset description, hardware and dependency versions.

Want to contribute a reproducible benchmark? Open a focused proposal in GitHub Issues.