Production-focused AWS learning track for designing, measuring, protecting, operating, and optimizing retrieval-augmented and agentic systems.
Define what retrieval can and cannot establish, then frame measurable boundaries before selecting architecture.
Build the smallest auditable AWS retrieval path with citations, explicit data boundaries, and useful failure signals.
Prepare, version, and ingest a corpus so retrieval quality has an inspectable foundation.
Understand representation, approximate-nearest-neighbor tradeoffs, and their effect on retrieval behavior.
Tune retrieval with testable relevance, citation, ranking, and context-selection criteria.
Diagnose retrieval failures with representative evaluation sets, traces, and controlled experiments.
Select AWS retrieval storage and indexing patterns against workload, operations, and governance needs.
Protect corpus, query, retrieval, model, and tool boundaries from data exposure and adversarial inputs.
Operate retrieval systems with traceable changes, meaningful signals, recovery plans, and safe rollouts.
Balance cost, latency, quality, and resource efficiency using workload-specific measurements.
Add planning and retrieval iteration only where it improves a verified user or operational outcome.
Connect governed tools and context through AgentCore and MCP with explicit authorization and audit boundaries.
Evaluate advanced retrieval patterns against evidence, complexity, and measurable workload benefit.
Validate production retrieval-system decisions across AWS architecture, quality, security, operations, and agentic retrieval.
Use these two primary sources as architecture-review guides while working through the track. They frame decisions across all six Well-Architected pillars; they are not certifications, service availability guarantees, or substitutes for workload evidence.
AWS guidance for designing, deploying, and operating generative AI workloads across operational excellence, security, reliability, performance efficiency, cost optimization, and sustainability.
Open the Generative AI LensAWS guidance for bounded autonomy, agent identity and tools, observability, evaluation, reliability, and cost-aware agent operations.
Open the Agentic AI Lens