Course Objective & Scope
LLMs do not know an organisation's internal policies, product manuals or contracts out of the box; Retrieval-Augmented Generation (RAG) is the foundational pattern for connecting models to enterprise knowledge with verifiable citations and without retraining. This course walks through both the ingestion pipeline and the runtime retrieval-and-generation loop from first principles.
Participants build a working RAG assistant over representative enterprise documents—covering chunking, embeddings, vector search and cited answer generation—while learning how to diagnose common failure modes.
Who Should Attend?
- Software developers starting internal knowledge-base and document Q&A projects
- Data scientists and AI engineers connecting LLM applications to enterprise data
- Solution architects and technical leads seeking a clear grasp of how RAG systems work
- Prerequisite: basic Python skills and familiarity with LLM API concepts
Key Takeaways
- Build an end-to-end RAG pipeline across ingestion, retrieval and answer generation
- Decide when to use RAG versus long-context prompting or fine-tuning
- Observe how document cleaning, chunk size and overlap affect retrieval quality
- Write grounded prompts that answer strictly from retrieved context and provide citations
- Distinguish retrieval failures from generation hallucinations during debugging
Curriculum
Why RAG? Architectural Overview
- Combining parametric model knowledge with external enterprise knowledge
- Components of the offline indexing and online query pipelines
Document Loading and Core Chunking
- Extracting clean text and metadata from PDF, Word, Markdown and HTML files
- Fixed-size and structure-aware (recursive/heading-based) chunking basics
Embedding Models and Vector Search
- Representing text in vector space and semantic search with cosine similarity
- Vector indexing and top-k nearest-neighbour retrieval
Prompt Assembly, Grounding and Citations
- Formatting retrieved chunks with metadata and source identifiers in the prompt
- Designing instructions that abstain when evidence is missing and emit inline citations
Common RAG Failure Modes and Baseline Quality Checks
- Separating missing retrieval, noisy context and generation hallucinations
- Running quick sanity checks with a seed Q&A test set
Hands-On: Building Your First Cited Enterprise RAG Assistant
- End-to-end RAG prototype over a sample corporate document collection
- Live comparison of different chunking and top-k settings
Other Courses in This Category
-
Deep dive into dense, sparse (SPLADE/BM25), Matryoshka and late-interaction (ColBERT) embeddings, Turkish/multilingual retrieval quality, domain adaptation and semantic search engineering.
- Live Course
- 6 modules
-
Engineering deep dive into ANN indexing (HNSW, IVF, DiskANN, PQ) and vector databases (pgvector, Milvus, Weaviate, FAISS, Pinecone), covering filtered search, multi-tenancy and capacity planning.
- Live Course
- 6 modules
-
Parsing complex PDFs, tables and scanned enterprise documents with Docling, Unstructured and VLMs, and applying structural, semantic, parent-child and contextual chunking for high-recall RAG.
- Live Course
- 6 modules
-
Engineering RAG systems for strict grounding, verifiable inline citations, conflict resolution across sources, calibrated abstention when evidence is missing and automated faithfulness verification.
- Live Course
- 6 modules
-
Advanced RAG
Enterprise RAG architectures covering BM25-dense hybrid retrieval, ColBERT, neural reranking, self-correcting Agentic RAG, GraphRAG, Text-to-SQL, and GPU-accelerated serving.
- Live Course
- 10 modules
-
Boosting RAG retrieval precision and recall by combining BM25/SPLADE lexical search with dense vectors via RRF, query rewriting (multi-query, HyDE) and cross-encoder or ColBERT reranking.
- Live Course
- 6 modules
Get in touch
