WEEK 5
Better RAG & Evaluation
A sharper assistant with a measured report.
Course Details
Students learn how to move beyond a basic RAG system by improving retrieval quality and measuring performance objectively. They implement advanced retrieval techniques such as HyDE and context engineering, create golden evaluation datasets, benchmark system performance before and after optimisation, and generate professional evaluation reports that quantify improvements instead of relying on subjective judgment.
WHAT YOU'LL LEARN
Query transformation (HyDE)
Context engineering
Building a golden test dataset
Before/after benchmarking
Turning results into a real repor
Topics Covered
Query transformation (HyDE)
Golden datasets
Context engineering
Before/after benchmarking
Project 5
RAGOptimizer · RAGBench
Project Details
Project 1: RAGOptimizer
Project Description
Upgrade the CitationRAG assistant by implementing advanced retrieval techniques, including HyDE query transformation and context engineering. Students optimize the retrieval pipeline, compare performance against the baseline system, and improve answer quality using evidence-driven evaluation.
Project 2: RAGBench
Project Description
Build a complete RAG evaluation framework that creates a golden test dataset, benchmarks system performance, compares before-and-after retrieval quality, and generates an automated evaluation report with key performance metrics.
Project Result
Project Result 1
Built an optimized Retrieval-Augmented Generation system that enhances retrieval accuracy through HyDE and context engineering. Evaluated improvements using benchmark datasets, measured retrieval quality, and demonstrated measurable gains over the original RAG pipeline.
Project Result 2
Developed an end-to-end RAG benchmarking platform capable of evaluating retrieval and answer quality using a golden dataset. Automated performance measurement, comparison reporting, and evaluation metrics to provide objective evidence of system improvements.
