Building Retrieval-Augmented Systems with Knowledge Graphs Course

Lesson 1 • Knowledge Graph Fundamentals

Covers nodes, edges, triples, and ontologies as building blocks of knowledge graphs. Provides the vocabulary needed for all subsequent graph construction topics.

Lesson 2 • Retrieval-Augmented Generation Overview

Introduces the RAG architecture: retriever, knowledge source, and generator components. Connects retrieval concepts to language model output quality.

Lesson 3 • Data Representation in Graph RAG

Compares RDF, labeled property graphs, and hypergraphs for RAG suitability. Guides students in selecting the right representation for their domain.

Lesson 1 • SPARQL for RDF Graphs

Covers SELECT, CONSTRUCT, and ASK queries against RDF triple stores. Equips students to retrieve semantic facts from ontology-backed knowledge graphs.

Lesson 2 • Graph Database Landscape

Surveys native graph databases, multi-model stores, and in-memory graph engines. Helps students match database capabilities to RAG system requirements.

Lesson 3 • Graph Schema Design Patterns

Presents reusable schema patterns such as event graphs, hierarchies, and bipartite graphs. Prepares students to design schemas that support efficient RAG retrieval.

Lesson 4 • Property Graph Query Language

Teaches pattern matching, filtering, and aggregation using a property graph query language. Directly enables fact retrieval queries used in later RAG integration chapters.

Lesson 1 • Ontology Alignment and Enrichment

Covers aligning domain ontologies, importing public knowledge bases, and enriching graph facts. Expands graph coverage to improve retrieval recall in RAG systems.

Lesson 2 • Information Extraction Pipelines

Covers named entity recognition, relation extraction, and coreference resolution from text. These extracted facts become the raw material for graph population.

Lesson 3 • Graph Quality Assurance

Introduces completeness, consistency, and accuracy metrics for knowledge graph validation. Quality gates here prevent retrieval errors in downstream RAG pipelines.

Lesson 4 • Entity Resolution and Deduplication

Addresses record linkage, fuzzy matching, and canonical entity merging across sources. Ensures graph integrity before retrieval queries are executed.

Lesson 1 • Graph Embedding Fundamentals

Explains translational, bilinear, and neural embedding families for knowledge graphs. Provides the theoretical grounding for choosing embeddings in RAG contexts.

Lesson 2 • Text-Graph Joint Embeddings

Teaches co-training text encoders with graph structure to align semantic and relational spaces. Enables unified retrieval across textual and graph knowledge sources.

Lesson 3 • Node and Subgraph Representation

Covers node2vec, GraphSAGE, and subgraph pooling for producing node-level and subgraph-level vectors. These vectors serve as retrieval units in graph-RAG systems.

Lesson 4 • Embedding Maintenance and Updates

Addresses incremental re-embedding, drift detection, and scheduled refresh strategies. Keeps retrieval quality stable as the underlying knowledge graph evolves.

Lesson 1 • Graph Traversal-Based Retrieval

Covers entity-anchored traversal, beam search over graphs, and path scoring for fact retrieval. Enables structured multi-hop reasoning unavailable in flat retrieval systems.

Lesson 2 • Retrieval Evaluation and Iteration

Introduces precision, recall, MRR, and NDCG for evaluating graph retrieval quality. Guides iterative improvement of retrieval components before generator integration.

Lesson 3 • Query Understanding and Decomposition

Teaches entity linking, relation detection, and multi-hop query decomposition from natural language. Converts user questions into structured graph retrieval operations.

Lesson 4 • Sparse and Dense Retrieval Baselines

Establishes BM25 and dense passage retrieval as baselines before introducing graph-aware methods. Provides benchmarks for measuring graph retrieval improvements.

Lesson 1 • Agentic Graph-RAG Architectures

Introduces tool-calling agents that iteratively query the knowledge graph during generation. Enables dynamic, multi-step reasoning over graph knowledge at inference time.

Lesson 2 • End-to-End Pipeline Assembly

Guides assembly of ingestion, retrieval, and generation components into a deployable pipeline. Prepares students for the system design and optimization chapters ahead.

Lesson 3 • Hallucination Reduction Techniques

Addresses attribution, fact verification, and confidence scoring to reduce unsupported claims. Ensures generated answers are traceable to specific graph facts.

Lesson 4 • Graph-Conditioned Generation

Teaches conditioning language models on graph-structured inputs using linearization and attention. Improves factual grounding beyond standard prompt-based injection.

Lesson 1 • Cost Management and Resource Planning

Covers compute, storage, and API cost modeling for graph-RAG infrastructure. Enables informed trade-off decisions between performance and operational budget.

Lesson 2 • Caching and Latency Optimization

Teaches query result caching, embedding cache layers, and request batching to reduce latency. Directly improves user-facing response times in production RAG systems.

Lesson 3 • Architecture Patterns for Graph-RAG

Presents microservice, monolithic, and event-driven architectures for graph-RAG deployments. Guides selection based on team size, query volume, and update frequency.

Lesson 4 • Scaling Graph Databases

Covers horizontal partitioning, replication, and read replica strategies for large knowledge graphs. Ensures retrieval throughput scales with growing graph size and query load.

Lesson 1 • Continuous Improvement Workflows

Presents regression testing, staged rollouts, and model refresh cadences for sustained quality. Operationalizes improvement as a repeatable engineering process.

Lesson 2 • Feedback Loops and Active Learning

Covers user feedback collection, implicit signal mining, and active learning for graph updates. Closes the loop between production usage and knowledge graph improvement.

Lesson 3 • End-to-End RAG Evaluation Frameworks

Covers faithfulness, answer relevance, context precision, and recall as holistic RAG metrics. Provides a complete evaluation toolkit applicable to graph-RAG pipelines.

Lesson 4 • Retrieval and Generation Observability

Introduces distributed tracing, retrieval logging, and generation audit trails for RAG systems. Enables root-cause analysis of quality regressions in production.