Global

Building reliable RAG and LLM applications: Grounding, evaluation and data leakage

A ready-made course you customise for your organisation, with a live session deck and two surveys. You only add the 6 details that are specific to your organisation.

See the live deckSee the full outline
Try the coursePreview the survey
Length
45 min
Lessons
7 lessons
Frequency
Before a team starts a first RAG or LLM project
Last reviewed
21 Oct 2026
Choosing a pattern and building the RAG pipeline
What would you do?

Retrieval returns near-duplicates from three document versions

Try it: this is one of 22 interactive activities in the course.

What's included

Three pieces that work together

Copy once, get everything. Learners prepare on their own, you answer questions live, and the surveys tell you what stuck.

Self-paced course

Short lessons with real workplace scenarios, not walls of text.

  • 7 lessons, about 45 minutes
  • Sorting activity, flashcards, timeline, accordion and quizzes
  • 10-question final check, 80% to pass

Live session deck

An AhaSlides presentation to run after the course, in person or online.

  • 10 slides
  • Quiz, poll, word cloud, scale slides
  • A closing slide for your contacts

Two surveys

Find out what landed, and whether people use it on the job.

  • Post-course feedback, about 2 minutes
  • 30-day check-in
  • Anonymous by default
  • Results feed your next refresher

Learning objectives

At the end of this course, learners can

Each objective and the lesson that covers it.

IDObjectiveSourceWhere it's covered
TO 127.0Achieve ≥ 80% on the final assessment-Whole course
SO 127.1Decide when RAG, long context, fine-tuning or a database query fits a problemBest practiceWhole course
SO 127.2Explain tokens, context windows, embeddings and retrieval well enough to debug a weak answerBest practiceWhole course
SO 127.3Design retrieval that respects document access permissionsOWASP Top 10 for LLM Apps 2025 LLM02Whole course
SO 127.4Prevent sensitive data disclosure and indirect prompt injection through retrieved contentOWASP Top 10 for LLM Apps 2025 LLM01, LLM02Whole course
SO 127.5Evaluate answers for grounding and faithfulness before release, and use the design review route + your infoNIST AI 600-1 GenAI Profile; organisation's review processWhole course

Your part: what a template can't do for you

  • Your workplace details: fill in the 6 highlighted fields so the course describes your own policies, contacts and locations.
  • Records: record completion in your own LMS or record system.

Course outline

7 lessons, about 45 minutes

Minutes per lesson are an estimate, shared out by page count.

ContentInteractiveVideoYou customise
1

Outdated answers and exposed documents: a real failure

3 pages

7 min
Before you beginNoteList
The policy assistantTextImageTimeline
What went wrongFlashcardsKey factQuiz
2

Tokens, context windows, embeddings and retrieval basics

3 pages

7 min
Tokens and contextAccordion
Embeddings and retrievalProcess
Match the termMatching quizFlashcards
3

Choosing a pattern and building the RAG pipeline

3 pages

7 min
Which pattern?Sorting activity
Order the pipelineSorting activity
Design choiceScenarioQuiz
4

Respecting document permissions during retrieval

3 pages

7 min
Where permissions breakTabs
Enforce at retrievalProcessYour info
CheckScenarioMulti-select quiz
5

Resisting hidden instructions in retrieved content

3 pages

7 min
Indirect prompt injectionTimeline
DefencesFlashcards
Spot the injectionSorting activityQuiz
6

Evaluation checklist and design review route

2 pages

5 min
EvaluationAccordion
Ungrounded or grounded?Sorting activityChecklistYour infoQuiz
7

Final assessment

Assessment

5 min
Final check10 questions80% to passCompletion page

Live deck and surveys

10 slides to run live, two surveys to close the loop

Pink slides are interactive. Both surveys are anonymous; pink questions are written for this course.

ContentTitle and agenda
PollHave you built a RAG or LLM application?
Word cloudWhat makes an LLM answer untrustworthy?
QuizWhich problem suits RAG best?
Content · customisePipeline on one slide (chunk → embed → retrieve → generate) and where permissions belong {{ACCESS_CONTROL_STANDARD}}
QuizA retrieved web page tells the model to email the user's notes. This is…
ScaleHow confident are you evaluating answer quality before release?
Open-endedHow we test answer quality today
Q&AAsk the trainer anything
Content · customiseClosing {{DESIGN_REVIEW_ROUTE}}, {{SURVEY_1_URL}}, {{SURVEY_2_URL}}
Post-course feedback: 10 questions, about 2 min
  1. Overall rating 5 stars
  2. Relevant, realistic, right length, helpful activities agree/disagree grid
  3. What was most useful? multi-select
  4. Confidence before this course, and now 1-5 scale
  5. Have you built a RAG or LLM application? multiple choice
  6. How do you currently evaluate answer quality? multiple choice
  7. How confident are you choosing between RAG, long context and a database query? 1-5 scale
  8. One thing we should change or add open text
  9. Would you recommend this course? NPS
30-day check-in
  1. Have you used what you learned since the course? multiple choice
  2. How confident are you today? Compare with day 1 1-5 scale
  3. A moment where the training helped, or where you weren't sure open text
  4. Which topics should the next refresher focus on? ranking

Results show whether people use the training on the job, not just whether they finished it.

Make it yours

6 fields to fill in, and you're ready

Everything else is done. These are the details only your organisation knows; each one is highlighted inside the course.

  • How document permissions are modelled{{ACCESS_CONTROL_STANDARD}}
  • How LLM features are reviewed before release{{DESIGN_REVIEW_ROUTE}}
  • Models and services engineers may use{{APPROVED_AI_TOOLS}}
  • Surveys and completion record{{SURVEY_1_URL}}
  • Surveys and completion record{{SURVEY_2_URL}}
  • Surveys and completion record{{LMS_OR_RECORD_SYSTEM}}

FAQ

Common questions

Can I run building reliable RAG and LLM applications: Grounding, evaluation and data leakage training online?

Yes. The course is self-paced and online, and the live deck lets you add a live session in the room or online.

How often is this training needed?

Before a team starts a first RAG or LLM project, and at each design review cycle.

Who needs this training?

Developers, data scientists and solution architects building LLM-based applications; run by engineering leads, data governance or L&D.

What records should I keep?

Record completion in your own LMS or record system, so you can show who finished the course and when.

Can I edit the content?

Yes, everything. Change the wording, swap scenarios for ones from your workplace, add your logo, or remove parts that don't apply.

Related templates

Often taught together

All AI skills templates →
Global

How generative AI works: Essentials for every employee

All staff · AI skills

30 min
Global

Finding the right work for AI: Spotting, testing and prioritising use cases in your team

All staff and team leads · AI skills

35 min
Global

Working smarter with generative AI: A safe everyday workflow for office tasks

All office staff · AI skills

40 min
Global

Copilot at work: Safe, effective use of your office-suite AI assistant

All staff licensed for Copilot or Copilot Chat · AI skills

40 min
Last reviewed: 21 October 2026Templates are a starting point, not legal advice.

Ready to run it?

Open the live deck to see how the session runs alongside the course and both surveys.

See the live deck