Structured Data Extraction with LLM JSON Schemas
How to use OpenAI's structured outputs to turn unstructured text into reliable database records — without parsing fragile prose.
Read more →Articles tagged "Architecture"
← All articlesBuilding Production Scrapers: Concurrency, Checkpoints, and Failure Recovery
How to build web scrapers that handle thousands of pages, recover from failures, and parallelize LLM calls without hitting rate limits.
Read more →LLM Provider Abstraction: Design for Swappability
Why wrapping LLM calls in provider classes is essential for production AI features — and how to design the abstraction layer.
Read more →Human-in-the-Loop ML: Building Quality Control Into AI Pipelines
Why every production ML system needs human oversight — and practical patterns for building it into your pipeline from day one.
Read more →Building an LLM Conversation Agent That Maintains Context
How to build an LLM agent that generates contextual messages by combining conversation history, user preferences, and available data.
Read more →Epic: From 317 Lines to 166 — A Controller Refactoring Journey
A four-part series documenting the systematic refactoring of a bloated admin controller into clean, reusable patterns.
Read more →Epic: From Scrape to Suggestion — Building an AI-Powered Recruiting Pipeline
A four-part series building an end-to-end AI recruiting pipeline — from scraping job listings to generating personalized candidate messages.
Read more →Zero-Downtime Table Renames on Heroku
Heroku's release phase creates a race condition during destructive migrations. Here's how to rename tables without errors using PostgreSQL views.
Read more →