AI curriculum

Course outline for Database Systems for AI Applications

An AI assistant may retrieve an outdated document even after the authoritative record changes. Compare stores for conversations, documents, relationships, and agent state using access patterns and consistency needs. Separate authoritative records from derived retrieval indexes and inspect concurrent updates and deletion propagation before adding more databases.

Technology: Data Engineering and Databases
Duration: 4 hours
Modules: 2 · 2 hours each
Level: Level 1

About Database Systems for AI Applications

Database Systems for AI Applications Course Objectives

Represent an AI assistant workload in relational, document, key-value, and graph models.
Compare conversation lookup, document retrieval, relationship queries, and agent-state updates.
Distinguish authoritative records from derived retrieval and embedding indexes.
Diagnose stale reads, concurrent agent updates, and incomplete deletion propagation.
Compare embedded and distributed responsibilities and justify a primary-store choice.

Pre-requisites

Basic application development and data modeling knowledge.
Familiarity with simple queries and reading small structured datasets.

Before this course

Complete the required courses or bring equivalent experience. Recommended courses add useful depth; they are not enrolment gates.

No prerequisite course is required. The entry skills above still apply.

Recommended preparation

Software Engineering Foundations for AI Engineers

Has basic application/data-modeling and query familiarity; this is the entry course for database model selection.

Find your place in the learning map

Lab Setup

Local SQLite or another embedded database, a scripting runtime, and supplied document/key-value/graph query fixtures.
Synthetic conversations, documents, agent-state records, concurrent-update traces, and mock retrieval-index correction/deletion fixtures.
No live model, vector-index implementation, managed subscription, or installation of four database servers is needed.

Detailed Course Outline

Module 1: Compare AI application access patterns

Relational databases: Represent assistant users, conversations, source records, and agent state with tables and keys; compare joins, integrity constraints, and concurrent updates.
Document databases: Store conversation messages and document metadata as structured documents; compare flexible access with duplicated-data update and deletion responsibilities.
Key-value databases: Use explicit keys for session lookup, caches, and checkpoints; distinguish temporary cached values from authoritative agent state and durable records.
Graph databases: Represent document, entity, and source relationships as connected records; compare traversal needs with relational joins before adding a separate graph store.

Practical outcome: Represent assistant conversations, source documents, relationships, and agent state in four models; compare lookup, joins/traversal, and update needs.

Module 2: Choose authoritative state and derived indexes

Embedded databases: Assess whether local storage meets query and consistency requirements while assigning backup, restoration, and maintenance responsibilities.
Distributed databases: Inspect concurrent agent updates and stale retrieval records; document consistency, correction, and deletion responsibilities when authoritative and derived data live separately.
Database selection: Choose stores for conversations, documents, relationships, and agent state; justify a derived retrieval index against synchronization, operating cost, and authoritative-record requirements.

Practical outcome: Inspect concurrent agent updates and stale retrieval traces, assign source-of-truth ownership, and define synchronization and deletion responsibilities for a supplied retrieval-index fixture.

Practical exercise

Produce four model sketches and execute or inspect three assistant-data queries in two supplied local representations.
Test a valid agent-state update and diagnose a lost update or stale read; state the required consistency behavior.
Trace a document correction and deletion across the authoritative store, conversation references, and a mock derived index; flag stale or retained copies.
Recommend one primary store and justify any separate retrieval index using measurable query, backup, consistency, and synchronization requirements.

What you will produce

Select stores for an AI application while separating authoritative records, agent state, and derived retrieval-index responsibilities.

How the work is checked

All four models preserve the workload's identities and relationship meanings.
Three query results agree across the two exercised representations.
Stale-read and lost-update fixtures are diagnosed correctly.
The decision explains authoritative state, correction/deletion propagation, consistency, maintenance, and derived-index synchronization costs.

Four-hour AI application data-model comparison with two exercised representations and curated distributed/index traces. Embeddings and retrieval indexes are introduced as derived data, not implemented. Detailed vector infrastructure follows later. Deleting a source record alone does not prove all derived copies were removed; query separation is not authorization.

When another approach is better

Do not add a graph or distributed store when simple relational queries meet the requirements.
Avoid a separate read model when synchronization and stale-data cost exceed its query benefit.

How we train

Ask about exercises, instructor feedback and the prior knowledge you need. Tool-specific courses marked provisional may become modules in a broader course.

Discuss your learning goals