Module 1: Design and develop Azure Cosmos DB database solutions
This learning path builds the skills to design and develop applications on Azure Cosmos DB for NoSQL. You start by evaluating whether the service fits an application, creating your first account and container, and configuring the throughput and consistency settings a workload needs.
You then connect from code: initializing the SDK with a client singleton, performing point operations and transactional batches, and writing queries whose request unit cost you can predict. With that foundation in place, you design a data model and partitioning strategy, which is the decision that most determines how an Azure Cosmos DB application scales.
The path closes with two capabilities that AI applications depend on. You process the change feed to build event-driven integrations and keep denormalized data in sync, then select and apply the AI-assisted development tools that connect an assistant to your data and give an agent durable memory.
Module 2: Secure, optimize, and deploy Azure Cosmos DB solutions
This learning path builds the skills to operate Azure Cosmos DB for NoSQL in production. It starts with the security surface: authenticating with Microsoft Entra ID instead of account keys, separating control-plane from data-plane permissions, restricting network access, and masking sensitive values in query results.
From there you protect the data and keep it available. You choose between periodic and continuous backup, recover from a mistake with a point-in-time restore, and design a multi-region topology, including the four failover mechanisms and the rules that decide which one you can run during an outage.
The middle of the path is performance. You design an indexing policy for the queries an application actually runs, then measure what those queries cost and reduce it, reading server-side metrics and index utilization rather than guessing. Monitoring follows: interpreting response codes and the four distinct causes of a 429, tracking latency against a service-level objective, and alerting on the signals that predict a problem.
The path closes at fleet scale, where you manage throughput and analytics across many Azure Cosmos DB accounts from one place instead of account by account.
Module 3: Implement AI and analytics capabilities in Azure Cosmos DB
This learning path builds the AI and analytics capabilities that distinguish Azure Cosmos DB for NoSQL from a general-purpose document store. You start with retrieval, configuring full-text policies and vector policies on a container, choosing a vector index type against its dimension and volume limits, and running each kind of search against real catalog data.
You then combine them. Hybrid search fuses keyword and semantic results with reciprocal rank fusion, and you learn to account for its cost separately across each stage, avoid the filter placement that quietly removes the results the semantic half exists to find, and measure relevance against a fixed evaluation set rather than by impression.
With retrieval working, you ground a language model on it. You size a retrieval unit, separate the text you embed from the text you send to the model, assemble a prompt with verifiable citations, and treat retrieved content as evidence rather than instruction, which is the trust boundary that keeps a retrieved document from rewriting the system prompt.
The path closes with the two capabilities that make an application durable and measurable. You design an agent memory store that separates short-lived conversation state from long-term facts a later session can find by meaning, and you mirror operational data into Microsoft Fabric OneLake so analysts can query it with T-SQL and Spark without consuming a request unit on the database that serves your users.