Tier 3 · 12 weeks · 18–22 total hours/week, including 6 live hours
Agentic AI & Scalable Systems
Build AI systems that retrieve information, use tools and complete bounded workflows while remaining inspectable, permission-aware and recoverable. Learn when an agent is warranted before learning how to add one.
Who it's for
Developers who already have structured LLM integration, asynchronous processing, validation and deployment experience.
Entry evidence
Tier 2 exit capability or equivalent: structured extraction, model error handling, durable jobs and basic evaluation. Full-stack experience alone does not meet this bar.
Capability at exit
Build and operate a bounded, permission-aware agentic system with measurable quality.
What you’ll build and learn
- Workflow versus agent: tool loops, stop conditions and failure budgets
- Tool design, permission boundaries and least privilege
- SQL and permission-aware retrieval, pgvector, hybrid search and agentic RAG
- Orchestration with LangGraph; bounded MCP interoperability
- Persistent state, approvals, streaming and interruption
- Tracing, evaluation, cost controls and adversarial testing
- An independent, non-trivial capstone with a named stakeholder
How you’ll show it
- A non-agent baseline and a justified architecture decision
- Tests proving unauthorised tools or resources fail
- Traces linking a user task to retrieval, tool calls and outcome
- An operational defence of the independent capstone
Credential: Agentic AI & Scalable Systems
Role direction: Applied AI, AI product, LLM or AI backend engineering — seniority remains experience-dependent.
Role readiness is assessed; it is not conferred by attendance.