{"id":8934,"date":"2026-10-23T10:29:59","date_gmt":"2026-10-23T02:29:59","guid":{"rendered":"https:\/\/drhariz.com\/blog\/?p=8934"},"modified":"2026-10-23T10:30:24","modified_gmt":"2026-10-23T02:30:24","slug":"ai-training-software-developers-malaysia","status":"publish","type":"post","link":"https:\/\/drhariz.com\/blog\/ai-training-software-developers-malaysia\/","title":{"rendered":"AI Training for Software Developers and IT Teams in Malaysia (2026): AI Coding Assistants, APIs and Secure Use"},"content":{"rendered":"\n

For AI<\/a> training for software developers and IT teams in Malaysia, Dr Muhamad Hariz Muhamad Adnan (drhariz.com), HRD Corp certified AI trainer and Senior Lecturer at UPSI<\/a>, runs hands-on programmes on AI coding assistants, LLM APIs, retrieval-augmented generation (RAG) and secure use. Sessions are built around your own stack and company policy, delivered on-site or live online in English or Bahasa Melayu.<\/strong><\/p>\n\n

Most Malaysian software teams already use an AI coding assistant, often without a shared policy, a review standard or any training. The result is uneven: a few developers get large productivity gains, others paste sensitive code into public tools, and team leads cannot tell whether AI-written code is being reviewed properly. This guide sets out what a developer-focused AI programme should cover, how it differs from general staff AI training, and how IT managers in Malaysia usually organise and fund it.<\/p>\n\n

Why developers need different AI training from other staff<\/h2>\n

General corporate AI workshops focus on writing emails, summarising reports and drafting presentations. That content is useful for office teams, but it does not help a backend engineer decide whether to trust a generated database migration, or a team lead set rules for AI-assisted pull requests. Generative AI<\/strong>, meaning models such as GPT, Gemini, Claude and Llama that produce text and code from a prompt, changes the daily work of developers in specific ways: how code is written, how it is reviewed, how applications are built on top of model APIs, and how company data is protected along the way.<\/p>\n

A developer programme therefore starts from engineering tasks. Participants work in their own languages and frameworks, on realistic repositories, and the hands-on time is spent on the decisions developers make every day.<\/p>\n\n

What a developer AI training programme should cover<\/h2>\n

Teams rarely need every module at once. Most IT managers pick two or three based on where the team is today. A broader view of course types by role is in Best AI Courses for Companies in Malaysia<\/a>, and a full module list is in the AI training syllabus for Malaysian companies<\/a>.<\/p>\n\n

1. AI coding assistants in daily work<\/h3>\n

Effective use of assistants such as GitHub Copilot, Gemini Code Assist, Claude and ChatGPT inside the IDE: writing clear instructions, giving the model the right context, generating tests first, refactoring legacy code, explaining unfamiliar codebases and writing documentation. The emphasis is on when the assistant helps and when it slows a developer down.<\/p>\n\n

2. Reviewing AI-generated code<\/h3>\n

AI-written code can look correct and still be wrong. This module covers review habits for AI-assisted pull requests: checking edge cases, verifying that suggested libraries and functions actually exist, spotting outdated APIs, and keeping test coverage meaningful. Team leads leave with a short review checklist they can add to their pull request template.<\/p>\n\n

3. LLM API fundamentals<\/h3>\n

For teams building AI features into products: calling model APIs, structuring prompts in code, handling tokens, cost and latency, getting structured output such as JSON, and evaluating responses so that features can be tested like any other code. Participants build a small working feature during the session.<\/p>\n\n

4. Retrieval-augmented generation (RAG)<\/h3>\n

Retrieval-augmented generation<\/strong> connects a language model to a company’s own documents so answers are grounded in internal knowledge rather than the model’s general training. The module covers chunking, embeddings, vector search and evaluation, with attention to Bahasa Melayu and mixed-language content, which many Malaysian organisations hold. Background reading: RAG for Bahasa Melayu<\/a>.<\/p>\n\n

5. Secure and responsible use<\/h3>\n

What must never be pasted into a public AI tool, how to treat secrets and credentials, risks such as prompt injection and insecure generated code, and how personal data obligations under Malaysia’s Personal Data Protection Act 2010 (PDPA) apply when customer data passes through AI systems. This module is often the one management asks for first.<\/p>\n\n

6. Team policy and workflow<\/h3>\n

Turning the above into a written team standard: approved tools, data classification rules, review requirements for AI-assisted code, and how to measure whether AI use is actually improving delivery. The output is a draft policy the IT manager can adapt and circulate.<\/p>\n\n

Who the programme is for<\/h2>\n