Multi-step agents, tool orchestration, memory, planning loops and human-in-the-loop controls — design agentic workflows Malaysian enterprises can audit, monitor and scale beyond one-off demos.
⏱Duration: 5 days / 40 hrs
💻Format: Instructor-Led + Agent Labs
🌐Delivery: Classroom · Virtual · Hybrid
✅Pass rate: 91%
📅Next intake: 21 Oct 2026
🤖
Agent patterns
ReAct, plan-and-execute, supervisor workers
🔧
Tool orchestration
APIs, functions, sandboxes and retries
🧠
Memory & state
Short-term context vs durable knowledge
👤
Human oversight
Approvals, escalation and audit trails
What this programme is
From chatbots to agentic systems.
Agentic AI moves beyond single-shot prompts into systems that plan, call tools and coordinate sub-tasks. This programme teaches architecture and implementation patterns for reliable agents — with emphasis on observability, cost control and governance for regulated Malaysian industries.
You will build and defend a multi-agent capstone that handles real enterprise constraints: retries, timeouts, human approval gates and structured logging.
Best taken after Gen AI or AI/ML Bootcamp. Graduates often pair with AI-102, LangGraph-style production stacks or internal platform engineering roles.
Who should take this course
💻
Senior developers
You ship backends and want agent architectures.
🧠
ML / Gen AI leads
You need patterns beyond basic RAG.
🏢
Platform teams
Building internal agent frameworks.
📊
Solution architects
Steering multi-agent rollouts.
🧪
Innovation labs
PoC → pilot with guardrails.
🔐
Regulated sectors
Banking, telco, GLC audit requirements.
Prerequisites
✓ Completed Gen AI or equivalent LLM project experience
✓ Comfortable with Python and REST APIs
✓ Laptop with 16 GB RAM recommended
→ Gen AI programme or prior RAG/tool-use project strongly recommended — ask enrolment if unsure.
Programme curriculum
Five days. Design → orchestrate → govern.
Morning architecture, afternoon implementation labs, capstone integration throughout the week.
Agent Labs
Orchestrated. Reviewed.
Labs use managed agent runtimes and sandbox APIs — production patterns without fragile local setup.
01
Tool agent
Reliable function calling with retries.
Tools
02
Multi-agent
Supervisor delegates sub-tasks.
Orchestration
03
Memory layer
Persist state across sessions.
State
04
Ops drill
Trace failures and tune prompts.
Observability
05
Capstone defense
Live multi-step workflow demo.
Demo
+ Mentor feedback on architecture docs, runbooks and test coverage.
Assessment
Capstone defense. Agent system artefact.
Passing requires working multi-step agent, observability hooks and live Q&A.
Capstone rubric
Agent flowMulti-step workflow with tool use
ReliabilityRetries, timeouts and error handling documented
GovernanceHuman approval or audit trail in place
PassingMeets rubric
ResitOne coached revision week
Our 3-Mock Exam Programme
01
Failure drill
Break the agent — then fix it.
02
Dry-run demo
Peer critique.
03
Steering Q&A
Cost and risk questions.
0%
Pass Rate
91% complete capstone first attempt.
We enforce operational discipline — agents that work in demos and under real failure modes.
Multi-agentObservabilityHITL91% passMentors
Why our pass rate is 91%
Prompt-only courses
No orchestration artefact.
Nexperts
Defended agent capstone + mentor sign-off.
Your agentic path
Build on Gen AI foundations.
Pair with Gen AI for full stack literacy, then AI-102 or platform engineering tracks.
