← Curriculum

G03 Stage C · Applied GenAI 48 h (16 T / 32 P)

AI Agents & Tool Orchestration

Builds controllable, recoverable agents that act safely on real systems.

Prerequisite
G02
Modules
5
NOS
Statutory mapping

Tech stack

LangGraphCrewAIMCPRedis

Modules

5 modules · 48 hours

M1 · Agent foundations

8 h
TOPICS
  • The agent loop
  • Reasoning and acting patterns
  • Planning versus reactive execution
  • When an agent is unnecessary
  • Cost and latency implications

Lab: Single-tool agent with explicit termination conditions

Course material for this module is in production.

M2 · Tool design

10 h
TOPICS
  • Tool schemas and descriptions
  • Function calling
  • Argument validation
  • Side-effect classification
  • Idempotency
  • Tool errors and recovery
  • Model Context Protocol servers

Lab: Build and expose a tool set including a stateful external action

Course material for this module is in production.

M3 · Orchestration

12 h
TOPICS
  • Graph and state-machine execution
  • Durable state and checkpointing
  • Resumability
  • Branching and conditional edges
  • Subgraphs
  • Multi-agent patterns: supervisor, hierarchical
  • Handoffs

Lab: Stateful multi-step workflow that resumes after induced failure

Course material for this module is in production.

M4 · Memory

8 h
TOPICS
  • Short-term context
  • Conversation summarisation
  • Long-term memory stores
  • Retrieval-backed memory
  • Memory hygiene and privacy

Lab: Add scoped, expiring memory to an assistant

Course material for this module is in production.

M5 · Control and safety

10 h
TOPICS
  • Human-in-the-loop approval gates
  • Permission scoping and least privilege for tools
  • Loop and cost caps
  • Timeouts
  • Sandboxing
  • Audit trails

Lab: Approval-gated agent with enforced spend and iteration ceilings

Course material for this module is in production.

Real-world work scenario

An agent authorised to issue refunds must never act alone above a threshold. Implement permission scoping, a human approval gate, idempotent tool calls, cost caps and a complete audit trail; then demonstrate correct behaviour under retry and partial-failure conditions.

Assessment

Stateful agent with approval gate and recovery behaviour