Testing LLMs With DeepEval, Promptfoo, RAG & CI/CD

Evaluate RAG apps, red-team for security, and automate quality gates in CI/CD - become job-ready in AI QA
What you'll learn

---

Explain why unit, integration, and E2E tests are not enough for generative AI
Test RAG end to end - retrieval quality and grounded generation along with agentic multi flow testing
Deepeval evaluation metrics , Promptfoo security testing
Put evaluations in CI/CD with thresholds, quality gates, and cost control
Requirements

---

Python programming expertise and Testing fundamentals
Description

---

AI/LLM Testing Mastery: From DeepEval to Production CI/CD
Traditional tests can be green while your chatbot is still wrong, biased, or jailbroken. This course teaches you how to test LLM applications the way production teams actually have to: with metrics, judges, adversarial checks, and quality gates.
You will learn to evaluate correctness, relevance, faithfulness, hallucination, toxicity, bias, and security. You will test RAG pipelines and AI agents, run prompt-injection and jailbreak cases, and wire the whole suite into GitHub Actions so a bad answer can block a deploy.
We use DeepEval and Promptfoo, local models with Ollama (no API key required to learn), and finish with a portfolio-ready capstone: a support chatbot, an evaluation suite, and a CI pipeline.
You will learn how to
Explain why unit, integration, and E2E tests are not enough for generative AI
Score free-form answers with LLM-as-judge instead of exact string matches
Test RAG end to end - retrieval quality and grounded generation
Test tool selection, multi-step agents, and memory
Red-team chatbots for injection, jailbreaks, and data leaks
Put evaluations in CI/CD with thresholds, quality gates, and cost control
Debug a failing eval: was the chatbot wrong, or was the judge wrong?
Includes hands-on labs, homework after each module, and a GitHub project.
Who this course is for

---

QA engineers, SDETs, and testers moving into AI quality roles
Homepage

---