M16 · Level III — AI Systems Engineering
Retrieval, RAG & Knowledge Systems
Engineer systems that retrieve external information and condition model computation on that information with measurable relevance, provenance and reliability.
Topics
Technical concepts
- 01Information Needs Before Models
- 02Inverted Index: The Classical Retrieval Primitive
- 03TF-IDF and Term Discrimination
- 04BM25: Saturation and Length Normalisation
- 05Query Processing and Lexical Normalisation
- 06Dense Retrieval: Embeddings as Search Representations
- 07Exact Vector Search
- 08Approximate Nearest Neighbour Retrieval
- 09Hybrid Retrieval: Lexical + Dense Evidence
- 10Metadata, Filters and Access Control
- 11Chunking: The Retrieval Unit Is a Model Choice
- 12Chunk Overlap and Boundary Effects
- 13Structured Data and Tables in Knowledge Systems
- 14Multimodal Retrieval
- 15Query Rewriting and Expansion
- 16Reranking: Spend More Compute on Fewer Candidates
- 17Retrieval Score Calibration
- 18Context Packing and Evidence Selection
- 19Retrieval-Augmented Generation: Complete Pipeline
- 20Grounding: Evidence Before Fluency
- 21Provenance, Citations and Source Traceability
- 22Freshness, Index Updates and Temporal Knowledge
- 23Conflicting and Inconsistent Sources
- 24Knowledge Graphs and Structured Relations
- 25Hybrid Knowledge: Graph + Text + Vectors
- 26Multimodal RAG
- 27Query Routing and Retrieval Planning
- 28Retrieval Security: Indirect Prompt Injection
- 29Retrieval Poisoning and Corpus Integrity
- 30RAG Evaluation: Retrieve, Ground, Answer
- 31RAG Evaluation Dataset Design
- 32Abstention and Insufficient Evidence
- 33Answer Verification and Claim-Level Checking
- 34Context Contamination and Irrelevant Evidence
- 35Caching and Knowledge-System State
- 36Observability and Provenance Logs
- 37Retrieval vs Tools: Where Knowledge Systems End
- 38Quantum Technology Bridge: Retrieval, Search and Knowledge
Practical implementation
- 01Laboratory I: Search Engine from First Principles
- 02Laboratory II: Exact Dense Retrieval
- 03Laboratory III: ANN Recall-Latency Trade-off
- 04Laboratory IV: Hybrid Retrieval and Reranking
- 05Laboratory V: End-to-End RAG
- 06Laboratory VI: Chunking and Context-Packing Study
- 07Laboratory VII: Graph + Text Knowledge System
- 08Laboratory VIII: RAG Security and Integrity
Tools and technologies
Project
Auditable knowledge system + evaluation harness
Expected learning outcomes
- 01Can implement and evaluate an inverted-index search system.
- 02Can implement a dense retrieval baseline and an ANN trade-off study.
- 03Can combine lexical and dense evidence and rerank candidates.
- 04Can engineer chunking and context selection under a declared token budget.
- 05Can build an auditable RAG pipeline with source provenance.
- 06Can evaluate grounding, abstention and conflict handling.
- 07Can enforce metadata and access-control boundaries before generation.
- 08Can diagnose prompt-injection and corpus-poisoning risks.
- 09Can reproduce a retrieval/RAG research result with documented deviations.
Assessment
- 01Daily practice
- 02Laboratories
- 03Research reproduction
- 04Integrated project
- 05Technical viva
- 06Engineering record
Source document: TUTOR Frontier Module 16 Retrieval RAG Knowledge Systems