AI Engineering:
Permanent Upskilling for Technical Staff
Build, evaluate, and deploy production AI agents using Google ADK. Model-agnostic across Gemini, OpenAI, and Claude. 3-day intensive with 8 hands-on labs covering agent fundamentals, multi-agent orchestration, RAG, MCP integration, and production deployment.
Workshop Dates
Register your interest and be the first to know when dates are announced.
Private Corporate Training
Looking to upskill your entire engineering or data team?
Exclusive sessions available for groups of 25-35 pax per class. Fully HRDC claimable.
8 Hands-On Labs
Production projects you will build and deploy during the 3-day intensive.
Lab 1: Hello World Agent
Build a minimal agent and swap across Gemini, OpenAI, Claude via LiteLLM.
Lab 2: Multi-Tool Agent
Weather, budget, web-search tools with safety guardrail callbacks.
Lab 3: RAG-Augmented Agent
Ingest documents, build embeddings, wire retrieval into ADK agent, measure quality.
Lab 4: Multi-Agent Orchestration
Movie-pitch team with Sequential, Parallel, and LoopAgent patterns.
Lab 5: Dynamic Agent Routing
AgentTools, root agent routing, workflow vs LLM-driven comparison.
Lab 6: MCP & Enterprise Integration
MCP server, tool filtering, database querying with realistic data.
Lab 7: Eval, Observability & Guardrails
LLM-as-judge, Langfuse tracing, cost dashboard, guardrail callbacks.
Lab 8: Capstone Build & Deploy
Full multi-agent system deployed to Cloud Run.
HRDC Training Architecture
A structured, hands-on 3-day program to master AI agent engineering for production systems.
Day 1: Agent Fundamentals, Tools & Retrieval (RAG)
The AI engineering paradigm shift, Google ADK architecture, and building your first production agents.
Core Theory
- AI Engineering Paradigm: AI Engineering vs Traditional Software — the paradigm shift, the observe-reason-act loop, and where agents deliver enterprise value.
- Google ADK & Model-Agnostic Design: ADK architecture: Agent, Tool, Runner, Session. Swap Gemini, OpenAI, Claude or open-weight models via LiteLLM. Live demo.
- RAG Architecture: Embeddings, chunking strategies, vector stores (Vertex AI, Chroma, pgvector). Hybrid search, re-ranking, retrieval evaluation.
- Prompt Engineering & Model Selection: System prompts, tool descriptions, output format constraints. Cost, latency and capability trade-offs across LLM providers.
Working Examples Built in Class
Build a minimal agent and swap it across Gemini, OpenAI, and Claude via LiteLLM to see model-agnostic design in action.
Weather, budget, and web-search tools with a before_model_callback safety guardrail blocking adversarial inputs.
Ingest a document corpus, build embeddings, wire a retrieval tool into an ADK agent, and measure retrieval quality.
Starter notebook stitching Labs 1, 2 & 4 — participants complete this before Day 1 to arrive ready to build.
Day 2: Multi-Agent Systems, Enterprise Integration & Model Customisation
Orchestrating multi-agent teams, connecting to enterprise systems via MCP, and understanding fine-tuning.
Core Theory
- Multi-Agent Orchestration: SequentialAgent, ParallelAgent, LoopAgent. Hierarchy, delegation, state sharing, and context passing between agents.
- MCP & Enterprise Integration: Model Context Protocol: connecting agents to databases, REST APIs via OpenAPI, and enterprise tools via OAuth + tool filtering.
- Model Customisation: Fine-Tuning, LoRA & MoE: The customisation ladder: prompting, RAG, LoRA/QLoRA, full fine-tuning. MoE models and the decision tree for when NOT to fine-tune.
- Agent-as-Tool & Dynamic Routing: Converting specialist agents into callable AgentTools. LLM-driven routing vs. deterministic workflow — when to use each.
Working Examples Built in Class
Movie-pitch team: Researcher, Writer, Critic. Build Sequential, then add Parallel critics and a LoopAgent revision cycle.
Convert specialists into AgentTools. Root agent dynamically routes queries. Compare workflow vs. LLM-driven routing.
Connect to an MCP server, implement tool filtering for safe access, query an external database with realistic data.
30-min concept-only session: LoRA, QLoRA, MoE architecture, and a decision framework for when NOT to fine-tune. No GPU needed.
Day 3: Production — Evaluation, Observability, Security & Deployment
Making agents production-ready with evaluation, tracing, guardrails, and Cloud Run deployment.
Core Theory
- Agent Evaluation & Testing: ADK evaluation framework, LLM-as-judge patterns, golden datasets, trajectory vs. response quality, and regression testing.
- Observability, Tracing & Cost Management: OpenTelemetry and Cloud Trace. Langfuse integration for tracing. Token and cost tracking per request. Latency budgets and quality alerts.
- Security, Guardrails & Responsible AI: Prompt injection defences, PII detection/redaction, tool-call authorisation, audit logging, and GDPR/EU AI Act compliance.
- Production Deployment: Containerise ADK agents. Deploy to Cloud Run (one command) and Vertex AI Agent Engine. Versioning, rollback, and monitoring.
Working Examples Built in Class
LLM-as-judge eval suite, Langfuse tracing, token-cost dashboard, prompt-injection and PII guardrail callbacks.
Full multi-agent system with tools, RAG, MCP, evaluation, guardrails, tracing — deployed to Cloud Run.
Teams demo their systems. Peer review and scoring: Best Architecture, Best Evaluation, Most Production-Ready.
Certificate of completion, curated resource guide, community links, self-study Colabs for fine-tuning, and Q&A.
Who Should Attend?
This intensive is designed for technical professionals who want to build production-grade AI systems.
Software Engineers & Developers
Engineers building AI-powered applications and integrating LLMs into existing systems.
Data Engineers & Data Scientists
Professionals designing data pipelines, analytics systems, and ML infrastructure.
Technical Leads & Solutions Architects
Leaders responsible for AI engineering strategy, architecture decisions, and team upskilling.
DevOps & Platform Engineers
Teams deploying, monitoring, and scaling AI systems in production environments.
Prerequisites
Experience the Workshop
Join a growing community of AI engineering professionals across Malaysia.
Our People
Learn from Malaysia's top AI engineering practitioners.
Dr Poo Kuan Hoong
Data Science, ML & AI Specialist
Deep expertise in AI/ML and Data Science platforms. Specialist in production-ready analytics solutions — spanning modern data engineering, MLOps, predictive workflows, and enterprise AI architecture. Advisor to national AI initiatives.
LinkedIn
Tze Jin
AI & ML Specialist
Deep expertise in machine learning and backend logic. Guides participants on integrating complex AI reasoning, database structures, model deployment, and building production-grade AI engineering systems.
LinkedInDetailed FAQ
Addressing your technical, logistical, and HRDC inquiries.
Course Fee
Transparent pricing for your AI engineering transformation.
Self-Funded (non-HRDC)
Kickstart your AI Engineering journey
- 3 full days of intensive training
- Complete course materials & templates
- Certificate of Completion
- 3-month post-training support
- Private community access
HRDC-Claimable
Upskill with your company's HRDC grant
- 3 full days of intensive training
- Complete course materials & templates
- Certificate of Completion
- 3-month post-training support
- Private community access
About AITraining2U
AITraining2U was established by professionals to close the divide between academic theory, business and practical industry demands. Our mission is to ensure that AI education translates directly into measurable, real-world results. Since 2025, we have upskilled over 1,200 professionals across Malaysia in AI, Business Transformation, Agentic Automation, and Vibe Coding.
Driven by a core philosophy of "100%-focus on success" our expert faculty delivers highly interactive, hands-on learning experiences focused entirely on implementation. We don't just teach prompt engineering; we teach you how to architect robust, autonomous systems.
Whether through bespoke corporate masterclasses or intensive public bootcamps, we actively partner with enterprise leaders, technical specialists, and government bodies to accelerate their digital transformation journey and build confident, AI-native organizations.