Python Developer Interview Prep Guide

The GIL ensures only one thread executes Python bytecode at a time, which simplifies memory management but limits CPU-bound parallelism. Use multiprocessing for CPU-bound work, asyncio for I/O-bound concurrency, and concurrent.futures for a high-level interface. Mention the PEP 703 free-threaded Python (no-GIL) experimental builds available in Python 3.13+ and their implications.

Use UploadFile for streaming file uploads, BackgroundTasks or Celery for async processing, and httpx for webhook delivery with retry logic. Show Pydantic models for request/response validation, dependency injection for database sessions, and proper error handling with custom exception handlers. Discuss how you would scale this with a task queue and separate worker processes.

Explain that decorators are higher-order functions that wrap other functions, using closures. Show the @functools.wraps pattern to preserve function metadata. Implement retry with configurable max_retries, base_delay, and exception types. Use time.sleep with exponential backoff (delay * 2^attempt) and optional jitter. Show both the function decorator and the decorator factory pattern.

Django provides batteries-included: ORM, admin, auth, migrations, and template engine. FastAPI offers async-first design, automatic OpenAPI docs, Pydantic validation, and superior performance for API-only services. Choose Django for full-stack web apps with admin needs; choose FastAPI for high-performance APIs, microservices, and ML model serving. Mention Django Ninja as a middle ground.

Profile with cProfile or py-spy first to find bottlenecks. Consider: replacing pandas with polars for faster DataFrame operations, using generators instead of loading everything into memory, parallelizing with multiprocessing or Dask, optimizing database queries with bulk operations, caching intermediate results, and using NumPy vectorization instead of Python loops. Provide specific speedup factors for each technique.

Show both the __enter__/__exit__ class approach and the contextlib.contextmanager decorator approach. For the connection pool, acquire a connection in __enter__, return it in __exit__ with proper cleanup on exceptions, and handle connection validation before returning to the pool. Discuss how contextlib.asynccontextmanager works for async code.

Discuss introducing type hints with mypy strict mode, setting up pre-commit hooks with ruff for linting and formatting, implementing pytest fixtures for clean test setup, adding CI/CD with coverage requirements, and creating shared utilities or internal packages. Quantify the impact: reduced bug rate, faster code reviews, or improved onboarding time.

Discuss uv or poetry for dependency resolution and lock files, virtual environments for isolation, pyproject.toml as the modern configuration standard, and Docker for reproducible deployments. Mention the difference between abstract dependencies (pyproject.toml) and concrete dependencies (lock file), and how to handle private package indexes.