CourseLarge Language Models: The Complete Course · Module 14: Applications, Patterns, and Frontiers · part 97 of 103
Part 97 · Module 14: Applications, Patterns, and Frontiers

Lesson 1: Seven proven patterns, one support desk

24 min read·13 Oct 2026
Module 14 · Lesson 1 · About 50 minutes · Needs: Module 6 Lesson 1 (structured output), Module 7 Lesson 2 (retrieval), Module 10 Lesson 3 (human review); the rest helps

The situation

It is Monday 31 August 2026, week 14. Drafted replies for billing and sign-in tickets are live behind a feature flag, and word has spread. Joel wants every refund ticket as a row his team can sort. Maya wants handover notes on long threads. Lena wants to ask the internal runbook questions. Tomasz wants a parser for invoice ids.

Each request is a pattern: a shape of LLM application that teams have run in production long enough to know where it breaks. You already own most of the parts. This lesson traces one ticket, T-1001 (Ana Duarte, a 288 USD duplicate charge), through seven of them.

who asks for pattern builds on Joel refund tickets as rows extraction schemas.py Maya handover notes summarisation schemas.py Lena drafts that cite grounded answers kb_search.py Tomasz an invoice-id parser code + verifier harness tests Maya replies a person signs review gate evals/ Joel a nightly table pipeline data.py: decide Lena ask the runbook private QA kb_search.py
Seven requests, seven proven patterns (amber, where the model works), and each leans on a part of supportdesk you already built (blue, your system).

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.