Beginner entry
Basic computer use and light command-line familiarity are enough. If Python is new, use the optional bridge first.
RAG & Knowledge Systems Engineering · RAG-KSE
Start with “what is RAG?” Finish by operating a trustworthy knowledge system.
Beginner-entry, engineer-exit. Learn each idea in plain English, build it, break it on purpose, diagnose the failing layer, then apply production controls for evidence, security, evaluation and operations.
My Learning progress
0% complete
0 of 18 curriculum modules completed · 0/18 currently published modules complete · 0 with lab + knowledge-check evidence.
Who this is for
No RAG, vector database or machine-learning background required.
The course keeps the entry barrier low without lowering the engineering exit standard.
Build progressively
One running “Evidence Desk” assistant grows across all 18 modules, so each new concept has context.
Engineer exit
Finish able to design, evaluate, diagnose, secure and operate retrieval-grounded knowledge systems with defensible evidence.
Learning contract
Understand → Build → Break → Diagnose → Engineer → Verify.
Every module repeats the same learning rhythm so a beginner always knows what to do next.
01UnderstandPlain-English mental model
02BuildSmallest working example
03BreakDeliberate failure case
04DiagnoseFind the failing pipeline layer
05EngineerProduction depth and trade-offs
06VerifyEvidence and release checklist
Four stages
Beginner → Builder → Engineer → Production.
Foundation
Understand RAG, build the smallest loop, preserve source identity and learn chunking without assuming ML knowledge.
Builder
Add embeddings, vector search, metadata filters, hybrid retrieval, reranking and safe query transformation.
Engineer
Engineer context, grounding, citations, source authority, evaluation and failure attribution.
Production
Operate freshness, security, bounded advanced retrieval and a release-gated production capstone.
Beginner bridge
Python you need—nothing more.
A real 45–60 minute bridge covers the exact Python patterns used by the course: variables, lists, dictionaries, functions, files, JSON, HTTP and virtual environments.
Concept first
Every technical term starts with a concrete mental model before implementation detail.
Open Python bridge →Deterministic baseline
Core labs use local Python and inspectable evidence before optional models, vector databases or framework abstractions.
Vocabulary support
Use the course glossary whenever terms such as embedding, top-k, reranker, provenance or faithfulness are unfamiliar.
Open glossary →Hands-on lab pack
Start deterministic. Then swap in real local models.
The downloadable pack keeps the required path Python-only and inspectable, then includes an optional Ollama practicum for real embeddings and local generation.
Deterministic core
13 runnable checks cover the course progression from evidence-first retrieval through security and release gating.
Download RAG lab packOptional Ollama practicum
Use a local embedding model and chat model through Ollama's REST API while preserving source-labelled evidence, citations and abstention.
Vendor neutral by design
Concepts are taught independently of a specific vector database or orchestration framework so the engineering transfers across stacks.
Failure map
Learn to locate the broken layer—not just tweak prompts.
Production RAG is a pipeline. A convincing answer can still be wrong because ingestion, retrieval, filtering, context, generation or citation failed.
Ingest→Retrieve→Filter / rerank→Context→Generate→Cite→Evaluate
Learning sequence
18 modules, one progressively engineered knowledge system.
Learner-facing titles explain the idea simply; engineering focus and stage show how depth increases.
Module 01Foundation
Why LLMs Need External Knowledge
Engineering focus: Why RAG exists
Understand why an LLM can sound confident yet still need external evidence, and learn the simplest mental model for retrieval-augmented generation.
Open module →Module 02Foundation
Your First RAG System
Engineering focus: Build the smallest working RAG loop
Build a tiny retrieval-and-generation flow so the architecture becomes concrete before introducing embeddings or vector databases.
Open module →Module 03Foundation
Documents, Parsing & Knowledge Sources
Engineering focus: Turn files into usable evidence
Learn how files become trustworthy knowledge objects with identity, text, metadata and parsing status.
Open module →Module 04Foundation
Chunking Made Understandable
Engineering focus: Split knowledge without losing meaning
Learn why chunking exists, how chunk size and overlap change retrieval, and how structure-aware chunking preserves context.
Open module →Module 05Builder
Embeddings: Intuition to Engineering
Engineering focus: Search by meaning, not only words
Understand embeddings conceptually first, then learn similarity, model identity and why embedding choice must be evaluated on your corpus.
Open module →Module 06Builder
Vector Search
Engineering focus: Retrieve semantically similar chunks
Build a small vector index, understand top-k retrieval, and learn why nearest-neighbour search needs filters and evaluation.
Open module →Module 07Builder
Metadata & Filtering
Engineering focus: Retrieve the right evidence for the right user
Use metadata to constrain retrieval by source, tenant, version, date, language or authorization boundary.
Open module →Module 08Builder
Keyword + Semantic Hybrid Search
Engineering focus: Combine exact words with meaning
Learn when lexical search and semantic search complement each other, and combine rankings without assuming one retriever is always best.
Open module →Module 09Builder
Reranking & Query Improvement
Engineering focus: Improve what reaches the model
Use candidate generation, reranking, query rewriting and decomposition to improve evidence quality without hiding the original question.
Open module →Module 10Engineer
Context Construction
Engineering focus: Build the evidence package the model actually sees
Select, order, deduplicate and budget retrieved evidence so the model receives useful context rather than a dump of search results.
Open module →Module 11Engineer
Grounding, Citations & Abstention
Engineering focus: Make answers show their evidence
Design outputs that distinguish supported answers from unsupported guesses and make citations useful for verification.
Open module →Module 12Engineer
Provenance, Authority & Conflicting Sources
Engineering focus: Know which source should win
Handle stale documents, superseded policies and conflicting evidence using explicit source authority and lifecycle metadata.
Open module →Module 13Engineer
RAG Evaluation from First Principles
Engineering focus: Measure retrieval and answer quality separately
Build a labelled evaluation set and use retrieval metrics plus grounded-answer checks so improvements are measurable rather than anecdotal.
Open module →Module 14Engineer
Diagnosing Retrieval & Generation Failures
Engineering focus: Find the layer that actually failed
Use traces and evidence inspection to distinguish ingestion, retrieval, reranking, context, generation and citation failures.
Open module →Module 15Production
Freshness & Knowledge Lifecycle
Engineering focus: Keep the index synchronized with reality
Design ingestion, updates, re-indexing, deletion and freshness checks so the knowledge system does not quietly serve stale evidence.
Open module →Module 16Production
Security, Prompt Injection & Access Control
Engineering focus: Treat retrieved content as hostile until proven otherwise
Protect the pipeline from malicious documents, cross-tenant leakage, unsafe tool influence and policy bypass.
Open module →Module 17Production
Advanced & Agentic Retrieval Patterns
Engineering focus: Use more steps only when the problem needs them
Learn routing, query decomposition, iterative retrieval and GraphRAG concepts without turning every RAG flow into an uncontrolled agent.
Open module →Module 18Production
Production Knowledge-System Capstone
Engineering focus: Build, evaluate, secure and operate the complete system
Integrate ingestion, retrieval, reranking, grounding, evaluation, security, freshness and observability into one production-shaped knowledge assistant.
Open module →Learning modules
FREE
All 18 modules and guided core labs are open learning content.
Assessment
FREE
30 randomized questions, 35-minute server timer, pass ≥75%, maximum 2 attempts per 24 hours; every attempt contains scenario/application questions.
Course completion certificate
FREE
Complete all modules, labs and checks to request the self-paced completion credential.
Request completion certificateAssessment-verified certificate
OPTIONAL · ₹499 incl. GST
Records successful GNU Group Learning Academy assessment completion. It is not vendor accreditation or a professional licence.
Reference anchors
Vendor-neutral engineering, aligned to recognized risk and evaluation practices.
NIST AI RMF / GenAI Profile
Use lifecycle risk discipline: govern, map, measure and manage rather than treating trust as a prompt-only concern.
NIST reference ↗OWASP GenAI Security
Treat prompt injection, poisoned knowledge, sensitive-data exposure and vector/embedding weaknesses as system threats.
OWASP reference ↗RAG evaluation
Measure retrieval and answer grounding separately; examples include Recall@k/MRR plus context precision/recall and faithfulness-style metrics.
Ragas reference ↗Advanced retrieval
Graph-based local/global retrieval is taught as an advanced pattern, not as a default replacement for simpler retrieval.
GraphRAG reference ↗Knowledge assessment
Prove that you can reason about the pipeline.
Questions emphasize architecture decisions, retrieval failures, evaluation interpretation, security and production operations—not acronym memorization.
84-question bank · 30-question attempt · 35 minutes · pass ≥75% · at least 14 scenarios/application questions.
18 MODULES→30 QUESTIONS · ≥14 SCENARIOS→35 MIN · PASS ≥75%→2 ATTEMPTS / 24H
Production capstone
Build and release-review an evidence-grounded knowledge assistant.
Ingest and version documents, retrieve with lexical + semantic signals, rerank, preserve provenance, cite sources, abstain when evidence is insufficient, test ACLs and prompt injection, evaluate retrieval and answer quality, exercise the HOLD path, and produce a promote/hold dossier.
Ingest→Retrieve→Rerank→Ground→Evaluate→Secure→Operate