02Learning path

RAG & Knowledge Systems

Build searchable, grounded knowledge assistants that answer from your documents and show their sources.

fig. — retrieval-augmented generation

Ingestion · offline

documents/
PDF, Markdown, HTML
parse + clean
keep headings & ids
chunk
structure-aware, overlap
embed
vector per chunk
index
vectors + metadata

shared index

Query · online

question
user input
retrieve top-k
hybrid: vector + keyword
rerank
keep the best 4
generate
answer only from sources
answer + [1][2]
citations to chunks

Ingestion runs offline when documents change; the query path runs on every question.

Stages: 5 stages
Suggested pace: Approx. 3–4 months part-time (suggested)
Programs: 3 in recommended order

About this path

Retrieval-augmented generation is the most common way to make an LLM useful on private or fast-changing information. This path covers each stage of a retrieval system in depth: ingestion, chunking, embeddings, search, reranking, and grounded generation with citations.

Every stage is treated as something you measure rather than guess at. You build a retrieval test set early and use it to compare chunking strategies, embedding models, and search configurations, so you can tell which changes actually improve results.

By the end you will have a knowledge assistant and a semantic search API, along with evaluation reports that document how you tuned them.

Who it’s forDevelopers building internal search, documentation assistants, or question-answering features over their own content.

Before you start

Comfortable writing Python
Basic SQL and familiarity with PostgreSQL
Experience calling a model API, or completion of the LLM Foundations program

Progression and milestones

01
Model and Embedding Basics
- Tokens and context budgets
- What embeddings represent
- Similarity measures
- Structured outputs for answers and citations
BuildEmbed a small document set and run nearest-neighbor queries from a Python script.
MilestoneYou can explain why two passages are or are not close in embedding space.
02
Ingestion and Chunking
- Parsing PDFs, HTML, and Markdown
- Preserving metadata and structure
- Fixed-size, recursive, and structure-aware chunking
- Incremental updates
BuildBuild an ingestion pipeline that loads a documentation set into PostgreSQL with pgvector.
MilestoneRe-running ingestion updates changed documents without duplicating chunks.
03
Search and Ranking
- Vector search and indexes
- BM25 keyword search
- Hybrid search with reciprocal rank fusion
- Cross-encoder reranking
- Metadata filtering
BuildExpose hybrid search with reranking through a FastAPI endpoint.
MilestoneHybrid search beats vector-only search on your retrieval test set, and you can show by how much.
04
Grounded Answers
- Context assembly within a token budget
- Prompting for grounded answers
- Structured citations
- Refusing when the answer is not in the corpus
BuildBuild a knowledge assistant that answers with citations and declines unanswerable questions.
MilestoneEvery answer either cites its sources or explicitly states that the information was not found.
05
Evaluation and Tuning
- Recall@k and MRR
- Groundedness and citation-accuracy checks
- Failure analysis by pipeline stage
- Retrieval regression tests
BuildWrite an evaluation report comparing two configurations of your pipeline.
MilestoneRetrieval and answer-quality checks run automatically on every change.

Recommended programs

Projects on this path

fig. — chunking with overlap

## 4. Refunds
Annual plans can be refunded within
30 days of purchase. Refunds go to
the original payment method.
### 4.1 Duplicate charges
If a card is charged twice, the

c-041 · lines 1–4
c-042 · lines 4–8
overlap · line 4

RAGDemo project

Semantic Search API

Build a search API that combines embeddings with keyword search, reranks results and serves them behind a FastAPI service with clear relevance metrics.

Embeddings
Hybrid search
Rank fusion
Reranking
Relevance metrics

View project brief

fig. — RAG query path

query
retrieve
rerank
generate
cite

RAGDemo project

Knowledge Assistant with RAG

Build an assistant that answers questions over a private document set, retrieves the passages it relies on and cites them in every answer. You own the full pipeline, from ingestion to source-aware responses.

Chunking
Embeddings
Vector search
Citations
Context budgets
Grounding

View project brief

By the end

Build ingestion pipelines that keep a search index in sync with changing documents
Choose and justify a chunking strategy using retrieval measurements
Implement hybrid search and reranking on PostgreSQL with pgvector
Generate grounded answers with accurate source citations
Evaluate retrieval and answer quality separately and diagnose which stage failed

Outcomes describe skills and portfolio artifacts. No employment outcome is promised.

Other learning paths

Move from experimenting with AI to engineering it.

Build the foundations, projects, and habits needed to design more dependable AI applications.

Explore Programs Browse Learning Paths

RAG & Knowledge Systems

About this path

Before you start

Progression and milestones

Model and Embedding Basics

Ingestion and Chunking

Search and Ranking

Grounded Answers

Evaluation and Tuning

Recommended programs

Projects on this path

Semantic Search API

Knowledge Assistant with RAG

By the end

Other learning paths

Move from experimenting with AI to engineering it.

RAG & Knowledge Systems

About this path

Before you start

Progression and milestones

Model and Embedding Basics

Ingestion and Chunking

Search and Ranking

Grounded Answers

Evaluation and Tuning

Recommended programs

Projects on this path

Semantic Search API

Knowledge Assistant with RAG

By the end

Other learning paths

Move from experimenting with AI to engineering it.