🚀
LLMsIntermediate4.5h Job-Ready
Building with Generative AI
Practical patterns to ship AI products
About this course
Move from playground prompts to production AI apps. RAG, evals, guardrails, and the realities of building with LLMs in 2026.
What you'll learn
- Architect a RAG pipeline end to end
- Run meaningful LLM evaluations
- Add guardrails and safety layers
- Estimate cost and latency
Curriculum
1
RAG, demystified
15 min
2
Embeddings and vector search
14 min
3
Evals: how to know your AI works
16 min
4
Cost, latency, and guardrails
12 min
Instructor · Tomás Reyes
Course Assignment
Ship a Tiny RAG With Evals
Production-shaped, not playground-shaped.
Brief
Build a minimum RAG over a small corpus (10-30 docs). Add a 20-row golden eval set. Report the current pass rate, p50 latency, and dollar cost per 100 queries. Propose one concrete change to improve the weakest number.
Deliverable
Repo or workflow + 20-row eval CSV + a one-page report with the three numbers.
Rubric
- RAG actually retrieves before generating (no faked retrieval)
- 20-row golden set is real and varied
- Pass rate, p50 latency, and cost are all measured (not estimated)
- Proposed improvement names a specific lever and a target number
- Sources are cited in every answer the system returns
Complete every lesson lab first — they're the raw material for this assignment.