AI · AIC-20

RAG & Knowledge Base

RAG & Knowledge Base is a course in building AI that answers from your organisation's own documents, covering embeddings, vector databases, chunking, hybrid search and reranking without deep programming. It suits IT, data, knowledge-management and helpdesk teams, and you leave with a RAG chatbot that cites its sources and reduces hallucination.

Updated 21 Aug 2026

From 8,910 THB / person 9,900 −10% excl. VAT 7% · group rates available

Corporate training quote Join the waiting list

PDFDownload the course outline

Duration12 hours · 2 days
FormatOnsite / live online
Next roundOn request
CertificateIncluded

Course overview

General AI models are strong at language but do not know your organization's internal information - manuals, policies, contracts, customer data or accumulated know-how. When asked something specific to the organization, AI often answers incorrectly (hallucination) or from generic, out-of-context information. This is a major barrier that keeps many organizations from using AI for work that demands high accuracy. RAG (Retrieval Augmented Generation) is the answer, connecting AI to the organization's knowledge base so it retrieves real data before every answer. The result is an AI assistant that answers accurately from the organization's documents, cites its sources, and greatly reduces hallucination. This course is designed for professionals and teams who want to actually build a RAG system,

focusing on business use rather than deep programming. Learners come to understand the RAG mechanism from document preparation, embeddings, storing in a vector database and retrieval, to having AI answer with citations. They learn advanced retrieval techniques such as hybrid search and reranking, how to measure and improve RAG quality, how to connect it to real systems via MCP and n8n, and the access control, security and PDPA practices needed for organizational data. All content is taught through workshops using realistic documents and problems, and finishes with a capstone project where learners build an end-to-end AI assistant that knows their own organization. (2 days, 6 hours per day, 12 hours in total, Intermediate to Advanced level, prior generative-AI use recommended, no programming required.)

What you’ll gain

Understand LLM limits (hallucination, knowledge cutoff, no internal data) and why organizations need RAG
Understand the full RAG mechanism: ingest, chunk, embed, store, retrieve, augment and generate
Understand embeddings and semantic search and choose a vector database that fits the organization
Prepare and chunk many document formats systematically and manage metadata
Build a knowledge base and retrieval system that answers from real data
Measure and improve RAG quality across retrieval and answer faithfulness
Use advanced retrieval techniques: hybrid search, reranking, query rewriting and metadata filtering
Build a RAG application or chatbot that answers with source citations and reduces hallucination
Connect RAG to real systems and data sources via MCP and n8n and keep the knowledge base current
Set access control, security and PDPA practices for organizational data in RAG

Who this course is for

IT, data and knowledge management teams that want to build search and Q&A over organizational data
Those who want to build an AI chatbot or assistant that answers accurately from internal documents
Customer service, support and helpdesk teams answering repetitive questions from manuals and knowledge bases
HR, legal, compliance and operations teams with many documents and policies
Business owners, executives and digital-transformation leads who want to bring organizational knowledge into AI

Prerequisites

Prior basic use of generative AI such as ChatGPT or Claude
General computer and internet skills
Understanding of your team's documents and knowledge to use as a case study during training
No programming background required, though reading JSON or having called an API helps you go deeper into vector databases and MCP

Curriculum

Course Details

A 2-day course, 6 hours per day (12 hours in total), delivered as a hands-on workshop with a capstone project. Intermediate to Advanced level. Learners take home their own RAG and knowledge base system, a RAG pipeline template, and an access control, security and PDPA checklist.

Day 1 Understanding RAG and Building a Knowledge Base

Section 1: Why Organizations Need RAG

LLM limits: hallucination, knowledge cutoff and not knowing internal data
What RAG is and how it turns a general AI into one that knows the organization
The difference between fine-tuning, putting data in the prompt, and doing RAG
RAG use cases: helpdesk, employee handbook, policies, contracts and knowledge bases

Section 2: Anatomy of RAG

The RAG pipeline: ingest, chunk, embed, store, retrieve, augment and generate
The indexing stage (prepared in advance) versus the query stage (answering at use time)
The role of each component and where quality commonly drops
An overview of market tools and choosing what fits the organization

Section 3: Embeddings and Vector Database

What embeddings are and how they enable semantic search
The difference between keyword search and semantic search
Choosing an embedding model and its effect on quality and cost
Vector databases (Qdrant, Pinecone, Chroma): how they work, choosing one, and self-hosted for internal data

Section 4: Document Preparation and Chunking

Ingesting many document formats: PDF, Word, Excel, web and databases
Chunking strategies: chunk size, overlap and splitting by document structure
The effect of chunking on retrieval quality and common problems
Managing metadata (source, date, access rights) for filtering and citation
Workshop: prepare and chunk a sample document set with metadata, ready to index

Section 5: Building a Knowledge Base and First Retrieval

Creating an index and loading embeddings into the vector database
Retrieval with similarity search and setting the number of results (top-k)
Reading retrieval results, judging relevance, and filtering by metadata
Workshop: build a knowledge base from the earlier workshop documents and test retrieval with real questions

Day 2 Quality, Advanced Retrieval, Real Use and Governance

Section 6: Measuring and Improving RAG Quality

Why RAG answers wrong: nothing retrieved, wrong retrieval, or the model ignoring the provided context
Two levels of quality: retrieval quality and answer faithfulness
Building an evaluation set from real work
How to trace causes and improve the pipeline point by point

Section 7: Advanced Retrieval

Hybrid search: combining keyword and semantic to cover both exact terms and meaning
Reranking: re-ordering retrieval results to better match the question
Query rewriting and multi-query for vague or broad questions
Metadata filtering and scoped search, such as by department or time period

Section 8: Building a RAG Application and Chatbot

Assembling retrieved context into the prompt and designing a system prompt for RAG
Forcing AI to answer with source citations and to say not found when it is not in the knowledge base
Handling multi-turn conversation and retrieval based on conversation context
Workshop: build a RAG chatbot that answers from the knowledge base and shows its sources

Section 9: Connecting RAG to Real Systems

Connecting real data sources via MCP (Model Context Protocol): Google Drive, SharePoint, Notion and databases
Setting a workflow to ingest documents and update the knowledge base automatically with n8n
Keeping the knowledge base current with re-indexing, and delivering RAG through real channels such as web, LINE or Microsoft Teams
Workshop: set a workflow to ingest new documents into the knowledge base automatically

Section 10: Security, Access Control and PDPA

Controlling access to knowledge base data (who can see which documents)
Classifying document confidentiality and preventing data leaks through AI answers
Personal data under PDPA and handling it before documents enter RAG
Cloud versus on-premise / self-hosted for sensitive data, and maintaining an audit trail

Section 11: Capstone Project and Real-world Adoption

Task: build an end-to-end AI assistant that knows the organization from a real document set
Prepare documents, index, improve retrieval quality and answer with citations, then test with real questions, measure quality and improve until usable
Present the work and get feedback, plus a first-30-days adoption plan and extending to enterprise knowledge and RAG-based AI agents

Schedule & training options

For individuals — public rounds

No public rounds are open right now. Join the waiting list and we will contact you first when the next round opens, or ask us on LINE. Or call 02-570-8449 or 088-807-9770

Join the waiting list Ask about rounds

For organisations — in-house / private

Tailor the content to your team’s tools and projects
Your dates, at your office or live online
Quotation with tax ID for procurement

Corporate training quote

Instructors

Samit KoyomFounder & CEO, IT Genius Engineering

Frequently asked questions

What is RAG, and how does it differ from fine-tuning a model?

RAG, retrieval-augmented generation, first searches your organisation's documents for relevant passages and then passes them to the LLM as material for its answer. Answers are therefore grounded in real documents, can cite their sources and update as soon as new documents are added. Fine-tuning trains the model further on your data, which suits adjusting style or output format but costs more and is a poor fit for information that changes often.

How should you chunk documents for RAG?

Chunking splits documents into pieces before embedding. Split along the document's structure, such as headings or paragraphs, rather than by character count alone. Choose a size that keeps enough context without mixing several topics, add a small overlap between chunks so key sentences are not cut, and attach metadata such as document title, section and date for filtering and citation. Then test with real questions and adjust the size.

How do hybrid search and reranking make RAG more accurate?

Semantic search is good at matching meaning but can miss exact terms such as product codes, names or clause numbers, which keyword search catches precisely. Hybrid search combines both, so you get meaning and exact matches. Reranking then uses a second model to read the question alongside each result and reorder them by true relevance, so the passages sent to the LLM are more on point and wrong answers become less likely.

Who is the RAG & Knowledge Base course for, and what should I know beforehand?

It suits IT, data and knowledge management teams building search and Q&A over company information, helpdesk and customer service teams answering repeated questions from manuals, and HR, legal or compliance teams with many policy documents. You should have used ChatGPT or Claude and know your own team's documents. Programming is not required, though reading JSON or having called an API helps with vector databases and MCP.

What will I be able to do after the RAG & Knowledge Base course?

You will be able to explain the RAG pipeline from ingest, chunk, embed and store through to retrieve and generate, choose an embedding model and vector database, prepare and chunk documents with metadata, apply hybrid search, reranking and query rewriting, measure retrieval quality and answer faithfulness, and build a chatbot that cites its sources, with access control in line with PDPA. Participants who complete the training receive a Certificate of Completion with its own number, which can be verified online.

Can the RAG & Knowledge Base course be run in-house for our team?

Yes. It can be run in-house for your team, with examples adapted to your own work, such as using your organisation's manuals, policies or knowledge base as the exercise material and setting document access rights to match your structure. You can request a quotation from this page, or contact us on LINE @itgenius or by phone on 02-570-8449.