LOCKEDIN / FORWARD-DEPLOYED ENGINEERING 101
Understand it.
Then build it.
You can be an experienced engineer and still be new to agents. Start with what each thing is and how it helps a person do useful work. Explore 64 original introductions across software, agents, cloud, security and delivery. Start here without a model key; each topic connects to deeper instruction and practice.
A place to start.
Choose a short reading route, or look up a concept below.
- 101UnderstandPlain language, visual walkthrough, vocabulary
- 102Build with guidanceOriginal lessons, examples and knowledge checks
- 103Practise and explainRunnable labs, failure cases and handover
Data access · RAG, hybrid search and graphs
Which retrieval approach fits the question?
Retrieval finds external evidence. RAG adds selected evidence to a model request. Keyword search, vector search, graph traversal and exact database queries are different ways to find information; none establishes permission or truth.
An FDE must match retrieval to the business question and maintain evidence through updates, permissions and failure.
UNDERSTAND THE SYSTEM
Find the right evidence before asking for an answer
A database query, text search, vector search and graph traversal answer different questions. Follow candidates through access, ranking and source verification.
Read the connections in plain English
- 01Question + trusted scope → Authoritative records
Use a scoped structured query when the question asks for an authoritative field.
- 02Authoritative records → Candidate retrieval
Ingestion, chunking and indexes run separately from answer generation.
- 03Question + trusted scope → Candidate retrieval
Search within the permitted corpus; relevance cannot override scope.
- 04Candidate retrieval → Access + freshness
Apply access before private passages enter a reranker or generator.
- 05Access + freshness → Selected evidence
Keep evidence provenance through context construction.
- 06Selected evidence → Answer or missing evidence
Check the answer against sources, not just its fluency.
Check your understanding: Does GraphRAG replace every vector index?
No. Graph-based retrieval is a design choice for relationship or corpus-level questions. It adds graph construction, provenance and update costs. Compare it with a simpler retrieval baseline on the same authorized questions.
CONNECT IT TO WHAT YOU KNOW
Start with something familiar.
Compare finding an exact record, looking up a phrase in an index and following references between documents. They are different search jobs.
Hybrid search combines search signals. Agentic retrieval changes workflow control. An index is not the source’s current permission policy.
Words worth knowing
- Hybrid search
- Combines retrieval signals, such as lexical and vector rankings.
- Reranker
- Reorders candidates; it cannot recover evidence absent from that set.
- GraphRAG
- Graph-related retrieval and synthesis; not an automatic replacement for simpler search.
- HNSW
- A graph-based approximate vector index, distinct from a business knowledge graph.
- Agentic retrieval
- A workflow in which an agent influences when or where to search.
Useful foundations
Open a definition and diagram here. Your lesson stays in place.
Data foundations · object storage, SQL and retrievalFiles, databases and search indexes: what goes where?
QUICK REFERENCE / 101
Files, databases and search indexes: what goes where?
An application needs somewhere to keep information and a way to find it again. Object storage keeps named objects such as documents and images. A relational database organizes records into tables you can query and update. A search index helps find relevant material. These are different jobs, even when one product can do more than one of them.
An FDE needs to explain where the customer's information lives before connecting it to an agent. A convincing answer is not enough if it used an outdated procedure, lost a ticket update or crossed a customer's access boundary.
An application pattern: retrieve, then generateWhat is retrieval-augmented generation?
QUICK REFERENCE / 101
What is retrieval-augmented generation?
Retrieval-augmented generation (RAG) is an application pattern. Your software finds relevant information and supplies it to a model before it answers. It is not a language, model or database product. It connects a model to a selected evidence collection without making that collection part of the model's learned parameters. Retrieval can use keyword search, semantic search or a combination.
Enterprise assistants need current company knowledge and clear boundaries on who may see it.
Numeric representations and similarity searchWhat are embeddings and vector search?
QUICK REFERENCE / 101
What are embeddings and vector search?
An embedding represents content as a list of numbers learned by a model. Vector search compares those representations to find similar items. It helps match meaning beyond exact words, but similarity is a retrieval signal, not a statement that two items are equally true or permitted.
FDEs need to diagnose retrieval quality instead of blaming every bad answer on the language model.
ONE EXAMPLE / FICTIONAL CASE
Follow the connections.
An exact ticket status needs a scoped record lookup. A paraphrased procedure may need semantic search. A relationship question may benefit from graph traversal.
UNDERSTAND THE SYSTEM
Find the right evidence before asking for an answer
A database query, text search, vector search and graph traversal answer different questions. Follow candidates through access, ranking and source verification.
See what moves and why.
Read the connections in plain English
- 01Question + trusted scope → Authoritative records
Use a scoped structured query when the question asks for an authoritative field.
- 02Authoritative records → Candidate retrieval
Ingestion, chunking and indexes run separately from answer generation.
- 03Question + trusted scope → Candidate retrieval
Search within the permitted corpus; relevance cannot override scope.
- 04Candidate retrieval → Access + freshness
Apply access before private passages enter a reranker or generator.
- 05Access + freshness → Selected evidence
Keep evidence provenance through context construction.
- 06Selected evidence → Answer or missing evidence
Check the answer against sources, not just its fluency.
Check your understanding: Does GraphRAG replace every vector index?
No. Graph-based retrieval is a design choice for relationship or corpus-level questions. It adds graph construction, provenance and update costs. Compare it with a simpler retrieval baseline on the same authorized questions.
Work through it, step by step.
Ask
Identify the answer type and authoritative source.
What you have now: A fact, passage or relationship.
Retrieve
Choose a simple baseline for candidate evidence.
What you have now: Scoped candidate IDs.
Check
Verify current access, source revision and relevance before downstream processing.
What you have now: Permitted current evidence.
Compare
Measure misses and unsupported answers on held-out cases.
What you have now: A reason to keep or change the approach.
FROM RECOGNITION TO REASONING
Explain the visual, predict a case and compare it with the linked workshop.
- 01
Name each component in plain English.
- 02
Predict the outcome before revealing the explanation.
- 03
Run the linked exercise and retain actual evidence with its limits.
A labeled explanation and a workshop observation you can reproduce.
Compare your reasoning when you are ready
Hybrid search combines search signals. Agentic retrieval changes workflow control. An index is not the source’s current permission policy.
Keep notes in your own working copy. Moving through these cards does not record completion.
TURN THE IDEA INTO USEFUL WORK
You can now go deeper.
Hybrid search combines search signals. Agentic retrieval changes workflow control. An index is not the source’s current permission policy.
Academy lessons use your existing training access. Free guides stay open.