#python

One pip install turns any of 1,404 LeetCode problems into a local practice environment: typed stubs, pytest suites, visual debugging, and full study plans.

The second bakefile in practice post: the helpers behind the patterns. Setup, secrets, publish, environments, config, the CLI, shipped once as library code.

Composition over configuration, Template Method, a uniform interface, workflows as libraries: four patterns from every repo I run bakefile in. Part one.

semantic-release versions main beautifully. Every other branch builds unversioned. zerv gives every commit a SemVer, PEP 440 included, without touching tags.

An audit of how five task runners share workflows between repos: Make include, Just import, Task includes, mise extends, Invoke collections. Where each breaks.

Every repo re-copies the same lint, test, and deploy recipes. Workflows need what code already has: inheritance, mixins, and versioned distribution.

Replace duplicated .env files with a layered Pydantic config framework: base, service, and instance mixins plus a CLI that unifies deploys across environments.

How a queue-driven API gateway with Cloud Tasks, Firestore caching, and queue-level rate limiting cut downtime by 70% and handled 40 req/s peaks.

A Python ML framework built on monadic design: DataPod state carriers, composable trained transformers, and reproducible pipelines from research to production.

How I replaced manual Excel-to-simulator data transfer with a Python tree-modeling tool for offshore pipeline design, cutting $49.8M from the investment plan.

A linear programming optimization app for a gas field of nearly 1,000 wells: the LP model, the data pipelines behind it, and the 15.2% production lift.