---

📅 Today’s Picks

Split Large Parquet Files Automatically with Polars

Problem

When writing large datasets to Parquet, you end up with either one massive file that is slow to read or must manually split data into smaller files.

Solution

With Polars PartitionMaxSize, output is automatically broken into multiple Parquet files according to a defined size limit.

This enables:

• Parallel reads across multiple cores
• Faster, more reliable cloud storage transfers

---

Coiled: One Decorator Replaces Your Entire Docker Workflow (Sponsored)

Problem

Have you ever had code work locally but fail on cloud VMs because of missing dependencies or version mismatches?

Docker solves this by freezing dependencies, but introduces friction: Dockerfiles, slow builds, registry pushes, and full redeploys for minor package changes.

Solution

Coiled can remove Docker from the workflow entirely. With a single decorator, it automatically syncs your local environment to the cloud.

Key features:

• Exact dependency replication from local to cloud
• No need for container builds or registry management
• Compatible with pandas, Polars, DuckDB, Dask, and more
• Faster deployments through smart caching

---

☕️ Weekly Finds

crewAI [LLM] – Framework for orchestrating role-playing autonomous AI agents that work together to accomplish complex tasks

Ray [MLOps] – Unified framework for scaling AI and Python applications from laptop to cluster with distributed runtime and ML libraries

Metabase [Data Viz] – Open-source business intelligence tool that lets everyone visualize, analyze, and share data insights

Looking for a specific tool? Explore 70+ Python tools →

---