Prerequisites Basic computer literacy and a willingness to document experiments.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Source permissions and document types Parsing and extraction quality Metadata and document versioning Deduplication and source identity RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools Document parser Python File corpus and manifest
Hands-on exercise Ingest a small approved corpus and record parse failures and source metadata. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project An ingestion pipeline with a source manifest and parse-quality report. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Every chunk can be traced to a permitted source document. Evidence must be reproducible and tied to RAG Systems.
Next: Chunking & metadata design
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Structural vs fixed-size chunking Overlap and semantic boundaries Metadata filters and access labels Language and table handling RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools Text splitter Document parser Metadata schema
Hands-on exercise Compare chunk strategies on the same question set and inspect context loss. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project A versioned chunking configuration with examples. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Chunk boundaries preserve meaning and support traceable citations. Evidence must be reproducible and tied to RAG Systems.
Next: Embeddings & vector indexing
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Embedding model selection Vector dimensions and similarity Index lifecycle and incremental updates Vector-store metadata filters RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools Embedding model Vector database Batch ingestion
Hands-on exercise Index a corpus and compare recall for exact and paraphrased questions. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project A repeatable vector index build and refresh job. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Index contents can be reproduced and updates do not create stale duplicates. Evidence must be reproducible and tied to RAG Systems.
Next: Retrieval, hybrid search & reranking
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Dense and lexical retrieval Hybrid fusion and metadata filtering Reranking and top-k selection Query rewriting trade-offs RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools Vector store BM25 search Reranker
Hands-on exercise Evaluate retrieval against a labelled set and measure recall@k and ranking quality. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project A retrieval service with documented tuning decisions. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Retrieval metrics improve against a recorded baseline. Evidence must be reproducible and tied to RAG Systems.
Next: Grounded generation & citation validation
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Context assembly and token budgeting Citation mapping and quote spans Abstention and insufficient context Answer-source consistency RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools LLM API RAG framework or custom pipeline Citation validator
Hands-on exercise Create questions with no answer in the corpus and verify the system abstains. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project A grounded Q&A endpoint with verifiable citations. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Each answer claim is supported or clearly marked as unknown. Evidence must be reproducible and tied to RAG Systems.
Next: Evaluation, security & access control
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Retrieval and answer evaluation Prompt injection in retrieved content Tenant/document-level permissions PII filtering and audit RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools RAG evaluation dataset Policy tests Redacted tracing
Hands-on exercise Attempt cross-role retrieval and poisoned-document prompt injection in a local test corpus. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project An evaluation and threat report with permission tests. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Users cannot retrieve documents outside their access scope. Evidence must be reproducible and tied to RAG Systems.
Next: Production refresh & observability
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Incremental refresh and tombstones Latency and cost telemetry Cache invalidation Index rollback and data retention RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools Scheduler or queue Monitoring dashboard Vector database admin
Hands-on exercise Update and remove source documents, then verify index consistency and cache behaviour. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project A monitored index refresh workflow and operator runbook. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria Changes propagate predictably and obsolete documents are not retrievable. Evidence must be reproducible and tied to RAG Systems.
Next: Capstone: auditable RAG system
✓ Mark this node as complete
Prerequisites Completion of the previous stage and a working local practice environment.
Learning objectives Explain the stage-specific model and its limits Apply the listed concepts using a controlled exercise Record results, failure cases and a reproducible evidence trail
Topics to learn Corpus governance Retrieval experiments Grounded generation and citations Operational and privacy readiness RAG Systems specialization: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Tools Parser and vector store Embedding model Evaluation harness
Hands-on exercise Demonstrate supported, unsupported and permission-restricted queries. Context for this route: RAG: focus on corpus ingestion, chunking, embeddings, hybrid retrieval, reranking, citations and access control.
Suggested project A deployable RAG product with a curated corpus, metrics, access controls and citations. The outcome must directly demonstrate RAG Systems, not merely list tools.
Learning resources Completion criteria A reviewer can reproduce metrics and trace answers to source records. Evidence must be reproducible and tied to RAG Systems.
End of roadmap: Complete the final project and build your portfolio.
✓ Mark this node as complete