Master data analysis and visualization with NumPy, Pandas, Matplotlib and Seaborn. Explore statistical methods, data preprocessing, and insights extraction techniques. Browse 34 curated articles covering data science with practical implementation detail.
Pandas, Polars, visualization, and preprocessing workflows
Performance, memory optimization, and reproducible analysis
Production-minded data pipelines beyond notebook prototypes
pola.rs made LazyFrame.collect streaming by default. Joins, group_by, and unpivot no longer keep incidental row order. Spill starts near 80% RAM with a 64 GB disk budget. CSV schemas now match by name.
DataCamp’s Oct 6 explainer puts TabPFN, TabICLv2, and TabFM on small random-split tables. An Oct 5 arXiv paper says activation alignment beats full-data XGBoost on 38 TabArena tasks. Your grouped split is the veto.
KDnuggets walked marimo as a Git-friendly notebook. CoreWeave put it in Forge on Oct 1. The graph is the product, not the banner.
Analytics Insight’s Sep 27 stack still opens on Python, SQL, NumPy, and pandas. DuckDB Lab’s Sep 26 note is partition pruning on v1.5.x. Those are not the same article.
DuckDB's Sep 22 post: dbt 2.0.0 bundles a pinned DuckDB over ADBC. catalogs.yml is v2-only. The bundled driver still will not load httpfs.
Talk Python 562 is Pedro Holanda on a lakehouse whose metadata is a database. QuestDB registers Parquet without copying it. You probably do not need a new platform.
A Latency Conference talk argued pandas should die. The useful part is Arrow: most tables never needed a cluster, and Polars 2.0 already made streaming the default.
Polars 2.0rc1 makes LazyFrame.collect() stream. Row order is no longer free. What to pin, what raises, and what to leave.
Data validation in Python has two dominant approaches: Pydantic for API and config validation, Pandera for DataFrame testing. Here's how they work, where they overlap, and why most teams need both.
Data cleaning is the least glamorous part of data science, but the right libraries turn it from a tedious chore into a satisfying pipeline. Here are five Python tools that make cleaning faster, more readable, and less error-prone.
Four Python libraries automate exploratory data analysis. One just got renamed, one focuses on speed, one on interactivity, and one on compact reports. Here's how to pick the right one.
Three libraries dominate Python data visualization, and each one excels in different scenarios. Here is a practical decision framework for choosing the right tool.
JetBrains and Microsoft both killed their notebook products in 2026, blaming AI for the shift. But Jupyter Notebook usage grew 75% in the same period. The real story is not about AI replacing notebooks — it's about which language owns the data science workflow.
You probably don't think about Apache Arrow. But every time you load a Parquet file in pandas, run a query in DuckDB, or pass data to Polars, Arrow is the reason it's fast. Here's what the 10-year-old project achieved — and what's coming next.
A practical decision guide for picking between Pandas, Polars, and DuckDB for your data workloads in 2026 — with real benchmarks, code comparisons, and honest tradeoffs.
Dirty data breaks pipelines, corrupts models, and wastes hours of debugging. Pandera and Great Expectations both solve data validation, but they take fundamentally different approaches. Here's how to choose.
How Python data engineers are adopting lakehouse-native patterns with DuckDB, Delta, and Iceberg to build self-healing pipelines that keep AI agents fed with live data.
DuckDB has become the go-to tool for Python data pipelines that need SQL-speed analytics without spinning up a server. Here's how to use it effectively.
Narwhals is a lightweight compatibility layer that lets you write DataFrame code once and execute it on pandas, Polars, cuDF, and more. Here's how to use it in production.
DuckDB brings fast SQL analytics directly into Python without a server. Learn how to use it for data analysis, Parquet queries, and workflows that beat pandas on large datasets.
A guide to modern geospatial analysis in Python, covering GeoPandas with Parquet, DuckDB's spatial extension, and cloud-native geospatial formats.
A practical migration guide for pandas 3.0, covering breaking changes, performance improvements from Arrow-native storage, and how to update legacy codebases.
How DuckDB's embedded OLAP engine is changing the way Python data teams handle analytical queries, eliminating the need for separate database servers in many workflows.
How Polars' lazy evaluation engine optimizes query plans, eliminates redundant work, and transforms data pipeline performance in Python.
An exploration of pandas' Arrow-backed data types and how they transform memory usage, I/O speed, and cross-language compatibility for data science workflows.
Polars 2.0 brings a redesigned API, GPU acceleration via cuDF integration, and streaming improvements that push DataFrame performance further. Here's how the update changes everyday data workflows in Python.
Complete guide to migrating from Pandas 2.x to 3.0. Learn about Copy-on-Write defaults, new string dtype, breaking changes, and step-by-step upgrade strategies.
A practical guide to Polars covering benchmarks, API comparisons, lazy evaluation, and when to migrate from Pandas. Includes real code examples and production patterns.
In-depth comparison of core data analysis tools like Pandas and NumPy, mastering the complete workflow from data cleaning to visualization, with practical code examples and modern best practices.
Master Pandas memory optimization with practical techniques. Reduce memory usage by 90%, process 10M-row datasets in seconds, and learn when to switch to Polars for massive workloads.
Learn how to build accurate time series forecasting models using Python's statsmodels library. Master ARIMA, SARIMA, and seasonal decomposition techniques.
Master data preprocessing techniques with Python. Learn to handle missing values, encode categories, scale features, and prepare datasets for machine learning with practical examples.
Learn how to create dynamic, web-based dashboards using Plotly Dash. This tutorial covers setup, callbacks, multi-page layouts, and deployment.
Learn how to create professional statistical visualizations in Python using Matplotlib and Seaborn. Step-by-step tutorial with real code examples.