This module develops the evaluative judgment needed to select the right Azure OpenAI model for a given enterprise use case—balancing capability, latency, cost, and context window requirements—and to design the complete deployment configuration that supports that selection. This is the primary architectural decision-making module in SC 4 and directly prepares learners for RAG architecture and agentic design work in subsequent courses.

This module develops the end-to-end skill of designing the retrieval layer of a RAG solution, from index schema and chunking strategy through embedding model selection and hybrid retrieval pipeline configuration, including the advanced production mechanics that distinguish an enterprise deployment from a proof-of-concept, using Azure AI Search as the primary retrieval service. To be able to complete this course the learner will require the minimum Azure AI search service tier of basic or better.

This module develops the skill of designing the agentic grounding layer of a RAG solution using Foundry IQ—covering the two-tier object model, output mode configuration, structural limits, identity-based permission enforcement, and multi-source data connectivity.

This module develops the skill of designing enterprise-grade system prompt architectures, moving beyond ad-hoc prompt writing to structured, governed, and maintainable prompt designs that enforce safety, grounding, and output consistency at the architectural level.

This module develops the skill of designing a structured evaluation framework for generative AI solutions, defining quality, safety, and groundedness metrics, configuring Microsoft Foundry evaluation flows, and producing the evaluation scorecard that serves as the ongoing quality signal for a production AI solution.

This module develops the strategic evaluation judgment needed to select the right build strategy, fine-tuning, RAG, or prompt engineering, for a given enterprise AI use case. Learners move from a five-criteria decision framework through realistic enterprise use case evaluation, producing the build strategy recommendation document that architects use to justify their choices to business stakeholders before a single training job is configured.

This module develops the end-to-end skill of designing a complete fine-tuning architecture in Microsoft Foundry, from dataset preparation through training configuration, evaluation integration, model version management, and model router design. Learners work through the pipeline decisions that determine whether a fine-tuned model reliably reaches production, and whether it remains reliable as usage patterns and data evolve.

Learn how to use generative AI tools to accelerate the production of architecture artifacts without compromising the judgment behind them. This module shows you how to use AI assistants to draft and critique system prompt components, generate evaluation metric definitions, representative evaluation datasets, edge-case test cases, and stress-test model selection rationale, making your architectural practice faster, more rigorous, and more defensible.

Learners produce a portfolio-ready Core AI Solution Architecture Design document for a provided enterprise scenario, integrating Azure OpenAI model selection and deployment design, RAG retrieval and grounding architecture, system prompt architecture, evaluation framework design, and build strategy justification into a single coherent solution architecture that mirrors the output a practicing Azure AI Solutions Architect would bring to a technical design review.