Understand How AI
Accelerators Actually Work

The GPU Fundamentals Reading Track suits university students and working professionals entering the machine-learning hardware space for the first time. No prior hardware background is assumed — the track starts from first principles and builds gradually. The Hands-On Lab Companion is better suited to developers or data analysts who already understand basic ML concepts and want to move into applied hardware reasoning.

No dedicated GPU is required. The reading track is entirely study-based. The lab companion uses publicly available hardware specification sheets and dashboard screenshots rather than requiring direct access to compute hardware. Learners who do have cloud GPU access can optionally apply concepts in a live environment, but it is not a prerequisite.

The Reading Track is structured across eight modules. Most learners work through one or two modules per week, completing the track in four to six weeks depending on reading pace. The Hands-On Lab Companion covers twelve sessions; a typical completion window is six to eight weeks. The Team Knowledge Programme spans multiple facilitated weeks, with a schedule agreed upon at the outset.

The programme includes facilitator-led sessions covering AI compute concepts and hardware terminology, a reference handbook tailored to your team's context, and follow-up Q&A time after each session. Materials are prepared with your organisation's typical workload types in mind, so discussions stay relevant to your daily engineering decisions rather than generic case studies.

The GPU Fundamentals Reading Track is priced at RM 490, the Hands-On Lab Companion at RM 1,850, and the Team Knowledge Programme at RM 4,600 for groups of up to fifteen. For team programmes, invoice-based payment with split terms is available upon request. Contact us to discuss what works for your organisation.

The curriculum draws primarily on NVIDIA GPU architecture as the dominant reference platform in AI compute today. Concepts such as CUDA core organisation, memory hierarchies, and throughput reasoning are explained through this lens. Transferable principles around parallelism and workload sizing apply broadly to other accelerator architectures, making the knowledge relevant beyond a single vendor.