https://datatweets.com/courses/generative-ai/shipping-ai-applications/Recent content in Shipping AI Applications on DATATWEETSHugoenCopyright (c) 2025 DatatweetsSun, 28 Jun 2026 09:00:00 +0200https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-1-from-prototype-to-production/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-1-from-prototype-to-production/A prototype handles the happy path; production handles everything else. Learn the five concerns that separate a script from a shipped product, and see the real token usage and cost numbers that production engineering is built on.https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-2-streaming-and-error-handling/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-2-streaming-and-error-handling/Reliability is the first production concern. Stream responses with the SDK’s streaming API, handle the anthropic exception hierarchy with try/except, and lean on built-in automatic retries with exponential backoff and timeouts.https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-3-managing-cost-and-tokens/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-3-managing-cost-and-tokens/Every token costs money. Read the usage object and convert it to dollars, estimate cost with count_tokens before sending, and pull the real cost levers — model choice, max_tokens, prompt trimming, and caching — while tracking spend in production.https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-4-securing-and-serving/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-4-securing-and-serving/Two production concerns at once: keep secrets safe (read keys from the environment, never hardcode) and serve your model behind a real API. Build a FastAPI endpoint with Pydantic schemas, input validation, and provider-error-to-HTTP mapping.https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-5-guided-project-deploying-an-ai-app/Fri, 14 Nov 2025 09:00:00 +0200https://datatweets.com/courses/generative-ai/shipping-ai-applications/lesson-5-guided-project-deploying-an-ai-app/The final capstone: build a production-shaped FastAPI service for Claude that applies all five concerns — reliability, cost, security, serving, and observability — with schemas, error mapping, cost-per-request, logging, and a health check, verified end to end.