Advanced RAG & Graph RAG

Live Online (VILT) & Classroom Corporate Training Course

Improve RAG systems beyond basic vector retrieval - advanced chunking, parent-child retrieval, hybrid search, query transformation, reranking, contextual retrieval, agentic RAG and Graph RAG using knowledge graphs and graph-based retrieval.

Expert-Led VILT & Classroom 16 Hours Level: Advanced Certificate of Completion
CloudLabs
Projects
Assessments
24/7 Support
Lifetime Access

Overview

Improve RAG systems beyond basic vector retrieval - advanced chunking, parent-child retrieval, hybrid search, query transformation, reranking, contextual retrieval, agentic RAG and Graph RAG using knowledge graphs and graph-based retrieval.

What You Will Learn

By the end of this course, learners will be able to:

  • Identify the limitations of basic vector-based RAG systems.
  • Apply advanced chunking and parent-child retrieval techniques.
  • Implement hybrid search using sparse and dense retrieval approaches such as BM25 and vector search.
  • Use query rewriting, query expansion and multi-query retrieval to improve recall.
  • Implement reranking and contextual retrieval to improve relevance.
  • Design controlled Agentic RAG workflows with iterative retrieval.
  • Explain Graph RAG, knowledge graphs, entities, relationships and graph-based retrieval.
  • Combine vector and graph retrieval for complex knowledge-intensive queries.
  • Evaluate advanced RAG systems using retrieval, context and answer-quality metrics.
  • Build and demonstrate an advanced RAG system incorporating multiple retrieval strategies.

Prerequisites

Working knowledge of Python, LLMs, embeddings, vector databases and basic RAG pipelines is recommended.

Course Outline

  • Review of the Basic RAG Pipeline
  • Common Retrieval Failures
  • Poor Chunk Boundaries
  • Missing Context and Context Fragmentation
  • Low Recall and Irrelevant Retrieval
  • Query-Document Vocabulary Mismatch
  • Long-Context and Context-Window Challenges
  • Hallucination and Grounding Issues
  • Latency and Cost Trade-offs
  • When Basic RAG is Not Enough

  • Limitations of Fixed-Size Chunking
  • Semantic Chunking
  • Structure-Aware Chunking
  • Hierarchical Document Representation
  • Parent-Child Chunking
  • Child-Chunk Retrieval and Parent-Context Expansion
  • Chunk Metadata and Document Hierarchy
  • Chunk Overlap Optimization
  • Selecting Chunking Strategies
  • Evaluating Chunk Quality

  • Dense vs Sparse Retrieval
  • Keyword Search Fundamentals
  • BM25 Retrieval
  • Vector Similarity Search
  • Vocabulary Mismatch and Semantic Matching
  • Combining BM25 and Vector Search
  • Reciprocal Rank Fusion
  • Weighted Retrieval Fusion
  • Hybrid Search Configuration
  • Evaluating Hybrid Retrieval

  • Why Query Transformation Matters
  • Query Rewriting
  • Query Normalization
  • Query Expansion
  • Adding Missing Context to Queries
  • Multi-Query Retrieval
  • Generating Query Variations
  • Handling Ambiguous Queries
  • Intent-Aware Retrieval
  • Query Transformation Evaluation

  • Why Initial Retrieval is Not Enough
  • Reranking Fundamentals
  • Cross-Encoder and Reranker Concepts
  • Candidate Generation and Reranking
  • Relevance Scoring
  • Contextual Retrieval
  • Enriching Chunks with Context
  • Selecting the Best Context for Generation
  • Reducing Irrelevant Context
  • Latency and Quality Trade-offs

  • What is Agentic RAG?
  • Basic RAG vs Agentic RAG
  • Retrieval as an Agent Tool
  • Retrieval Planning
  • Iterative and Adaptive Retrieval
  • Query Generation by Agents
  • Tool Calling for Search and Retrieval
  • Evaluating Intermediate Retrieval Results
  • Handling Retrieval Failures
  • Designing Controlled Agentic RAG Workflows

  • Graph RAG Fundamentals
  • Why Use Graphs for Retrieval?
  • Vector RAG vs Graph RAG
  • Knowledge Graph Concepts
  • Entities, Nodes and Relationships
  • Properties and Graph Schema
  • Knowledge Graph Construction
  • Entity Extraction and Relationship Extraction
  • Graph Storage and Querying
  • Graph-Based Retrieval
  • Neighborhood and Subgraph Retrieval
  • Combining Graph and Vector Retrieval
  • Graph Context Construction
  • Graph RAG Use Cases

  • Evaluation Challenges in Advanced RAG
  • Retrieval Precision and Recall
  • Top-k and Ranking Quality
  • Query Transformation Evaluation
  • Reranking Evaluation
  • Context Relevance
  • Answer Relevance
  • Faithfulness and Groundedness
  • Graph Retrieval Evaluation
  • End-to-End RAG Evaluation
  • Building an Advanced RAG Evaluation Dataset

  • Define the Advanced RAG Use Case
  • Design the End-to-End Architecture
  • Prepare and Ingest the Knowledge Base
  • Implement Advanced Semantic Chunking
  • Implement Parent-Child Retrieval
  • Generate and Store Embeddings
  • Configure Vector Search
  • Implement BM25 Retrieval
  • Combine Sparse and Dense Retrieval
  • Implement Query Rewriting
  • Implement Multi-Query Retrieval
  • Add a Reranking Stage
  • Implement Contextual Retrieval
  • Design an Agentic Retrieval Workflow
  • Extract Entities and Relationships
  • Build a Knowledge Graph
  • Implement Graph-Based Retrieval
  • Combine Vector and Graph Retrieval
  • Construct the Final RAG Context
  • Generate Grounded Responses
  • Add Source Attribution
  • Build an Evaluation Dataset
  • Evaluate Retrieval and Answer Quality
  • Optimize Accuracy, Latency and Cost
  • Final Demonstration

Available Training Modes

Pick the format that fits your team.

Same authorised curriculum, same trainers, same hands-on cloud labs — delivered the way that works for you.

Live Online (VILT)

Real-time instructor-led sessions over Zoom or Teams. Same classroom, different time zones.

Most popular

Classroom

Face-to-face training delivered at your office, our Bengaluru centre, or any partner venue worldwide.

Onsite

Self-Paced

Recorded sessions plus 24/7 access to cloud labs and assessments. Learn at the pace that works for each engineer.

On-demand

Blended

Live workshops with self-paced reinforcement and project-based labs. Best for hybrid teams across regions.

Hybrid teams
All modes include: hands-on cloud labs, recordings, assessments, certificate of completion. Talk to a solutions advisor →

Our Training Process

How a course becomes measurable skill.

One contract, five steps, zero handoffs. From discovery to deployment, the same Synergific team owns the outcome — not a chain of vendors.

5 Steps from your scoping call to certified, productive engineers.
01

Discover & set goals

We start with a scoping call to understand your team's current skill level, target outcomes, deadlines, and certification needs — then translate that into a measurable success plan with named owners on both sides.

02

Curate the right path

We map the optimal learning path — instructor-led, self-paced, or blended — with hands-on cloud labs, prerequisite refreshers, and certification vouchers built in. No filler modules, no padded curriculum.

03

Deliver hands-on training

Authorised trainers run live sessions backed by 24/7 cloud labs and real-world projects. Theory and practice on the same day — learners stop forgetting concepts before they get to apply them.

04

Assess & mentor

Continuous skill checks, mock exams, and 1:1 mentoring keep the program honest. If anyone falls behind, we course-correct in-flight — you'll never find out at the end that two engineers couldn't keep up.

05

Certify & apply on the job

Voucher-backed certification, post-training office hours, and 30-day reinforcement so skills land on real work — not just on the exam scorecard. Success measured after the course ends, not before.

WhatsApp +91 9541 551 557
New · Limited window Try any lab free for a day