09 Architecting LLM & Agentic Applications

Architect a production-grade deep-research agent from first principles —owning the agent loop, tool surface, retrieval strategy, and verification layer yourself. Each layer is motivated by watching the simpler version fail, so you gain the architectural judgment to design agentic systems for your own problems.

Advanced    Architecture Workshop

What you´ll learn

• Agent loop and a disciplined tool surface; query decomposition, planning, and query generation/rewriting

• Retrieval strategy: search-API trade-offs (keyword vs. semantic vs. freshness), reranking, and the live-fetch-vs-pre-crawl decision

• Context engineering and evidence memory under a token budget; single-agent vs. orchestrator-worker multi-agent designs

• Grounding and citation discipline (“no citation, no claim”); anti-hallucination: claim decomposition, NLI checks, runtime judge,self-repair

• Eval-driven development and production hardening: LLM-as-judge, faithfulness metrics, tracing, cost/latency budgets, prompt-injection guardrails

Pick the path that matches your team – Read more

Target audience

Mid-to-Senior Software Engineers —New to Agents

Prerequisites

Comfort with Python and APIs. No prior LLM-engineering experience required.

 

Other relevant courses