https://renezander.com/tags/llm-engineering/Recent content in Llm-Engineering on René Zander | AI Automation ConsultantHugoenFri, 10 Apr 2026 08:00:00 +0200https://renezander.com/guides/production-ai-agent-architecture/Fri, 10 Apr 2026 08:00:00 +0200https://renezander.com/guides/production-ai-agent-architecture/<p>Most agent tutorials end at &ldquo;the model calls tools in a loop, done.&rdquo; That works for a demo. It falls apart the first time a tool 500s, a user asks something off-script, or the token bill crosses $20 on a single task. Production AI agent architecture is the set of patterns that keep that loop alive when reality hits.</p> <p>I run 10 agents in production right now. Bash scripts calling <code>claude -p</code>, scheduled via systemd timers, reporting outcomes to Telegram. Not fancy. They ship work every day because the architecture around the loop is boring and deliberate. This guide is that playbook: the patterns, the must-haves, the anti-patterns, and the opinionated verdicts on what to use when.</p>https://renezander.com/guides/migrate-openai-to-claude/Sat, 04 Apr 2026 10:00:00 +0200https://renezander.com/guides/migrate-openai-to-claude/<p>Most teams I talk to arrive at the same moment: the OpenAI bill crosses $500/month, an agent loop that worked on GPT-4o starts fumbling tool calls, or legal raises an eyebrow about single-provider risk. Then the question lands in my inbox: what does it actually take to migrate OpenAI to Claude?</p> <p>Short answer: a weekend if you have one endpoint, two weeks if you have a real product. The SDKs are similar enough that the ported code looks boring. The interesting work is in the prompts, the tool use loop, and the parts of your codebase that silently depend on OpenAI-specific behavior like <code>seed</code>, <code>logprobs</code>, or the <code>response_format</code> JSON schema flag.</p>