Scaling Your Pandas Analytics With Modin

on YouTube ↗21m

" summary

Ponder offers Modin as a drop-in replacement for pandas that parallelizes DataFrame ops (supports Dask and Ray backends and single-node use), claiming orders-of-magnitude speedups on a 2 GB New York taxi example and enabling a finance pipeline to process 16 months of time-series data instead of 6 months. Modin reports >4M downloads, adoption by >10% of Fortune 100, and API compatibility so most workflows run with a single line import change to Modin while keeping existing pandas code.

AI-written summary · about

" chapters (10)

1. 0:00Introduction
2. 2:41What is pandas
3. 4:12Why pandas is so awesome
4. 5:50Pandas limitations
5. 9:01Modin
6. 10:53Modin Demo
7. 12:10Why Modin
8. 13:33Case Study
9. 15:57Recap
10. 18:58Modin vs DataFrame

" related talks

• 21mScaling Interactive Data Science with Modin and RayDevin Petersohn · Ray Summit 2021
• 26mBodo: Supercomputing-Like Performance and Scale for Python/PandasEhsan Totoni · XtremePython 2021
• 32mPandas, Polars and the DataFrame Consortium.Alessandro Romano · PyCon Italia 2024
• 30mIs the great dataframe showdown finally over? Enter: polarsLuca Baggi · PyCon Italia 2023
• 06mPython Data Scientists 10x FasterShashank Yadav · PyCon India 2020
• 24mEnd-to-End Computation on the GPU with a GPU Data FrameTodd Mostak · Scale By the Bay 2017
• 23mAccelerate the Machine Learning Pipeline on Very Large DatasetsAaron Williams · O’Reilly Jupyter Pop-up Boston 2018
• 58mFrom Pandas to Polars Upgrading Your Data WorkflowMatt Harrison · PyBay 2024