Description:

Many teams can build a basic Retrieval-Augmented Generation (RAG) prototype - but far fewer can make it accurate, reliable, and production-ready.

In this focused 90-minute advanced workshop, participants will learn the practical techniques used to transform a simple RAG pipeline into a high-performance AI system suitable for real-world deployment. The session concentrates on the three pillars of effective RAG design:

1. Intelligent document chunking
2. Advanced retrieval strategies
3. Automated evaluation and quality control

By the end of the session, participants will walk away with clear reference architecture, optimization techniques, and a practical production-readiness checklist that can immediately improve their existing no-code or low-code AI workflow

Duration:

Half Day

Course Code: BDT 541

Learning Objectives:

After this course, you will be able to:

1. Identify the top 3 failure modes in RAG pipelines
2. Select optimal chunking and metadata strategies for different document types
3. Implement Hybrid Search and Reranking to improve retrieval precision
4. Apply basic "Groundedness" evaluations to reduce hallucinations

AI Practitioners, Solution Architects, Technical Product Managers, Data Teams, Developers, and automation specialists building or deploying RAG-powered applications.

Participants should have a foundational understanding of Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) concepts

Course Outline:

1. The RAG Reality Check: Architecture & Failures
   1. The "Naïve RAG" vs. "Advanced RAG" gap: Why simple setups fail in production
   2. The Triad of Failure: Low Faithfulness (hallucinations), Low Relevance (noise), and Low Recall (missing info)
   3. Quick Audit: Identifying where your current pipeline is leaking value

2. Data Prep: Precision Chunking & Metadata
   1. Beyond Fixed-Size: Moving to semantic boundaries and recursive character splitting
   2. The Context Window vs. Retrieval Granularity: Finding the "Goldilocks" chunk size
   3. Metadata Injection: Using "Self-Querying" and filters to bypass vector noise
   4. Rapid Lab: Visualizing how different chunk sizes impact a single document's retrieval

3. Retrieval Mastery: Hybrid Search & Reranking
   1. Why Vector Search isn't enough: The case for Hybrid Search (Keyword + Semantic)
   2. The Power of Reranking: Why "Top K" is often wrong and how a second pass reranked fixes it
   3. Similarity Thresholds: Keeping the "garbage" out of your LLM context

4. Grounding & Prompt Guardrails
   1. The "Reference" Mandate: Structuring prompts to force citations
   2. Handling "I don't know": Preventing hallucinations on out-of-scope queries
   3. Context Injection: Efficiently formatting retrieved data for the LLM

5. Evaluation & The Production Checklist
   1. The RAG Triad Metrics: Faithfulness, Answer Relevance, and Context Precision
   2. LLM-as-a-Judge: Using automated tools to grade your RAG performance

Training material provided: Yes (Digital format)

Hands-on Lab: Students will be provided with docker compose file and n8n workflow JSON.