Announcing the AMD ROCm Certification Program — Your Pathway to AMD GPU Excellence

The AI and high-performance computing (HPC) landscape is scaling fast, and teams need practitioners who can build, optimize, and operate on AMD GPUs with confidence. The new ROCm™ Certification Program meets that need with a clear, hands-on pathway for developers and platform engineers to prove their skills across the AMD open software stack.

The Associate level certification launches on July 24, 2026.

Why this certification matters

• Industry demand: Organizations are adopting multi-GPU and multi-node clusters to train and serve AI models cost-effectively. Validated skills on AMD GPUs are increasingly sought after.
• Open ecosystem: ROCm software and HIP offer an open, performant stack for AI and HPC. This program validates real, practical ability using those tools.
• Hands-on credibility: Each level emphasizes applied skills, not just theory; so your certification reflects what you can do in production.

Program overview

Level 1: Associate (launching July 24, 2026) Validate your expertise in AMD accelerated computing. The Associate certification equips developers with the knowledge and hands-on skills to design, optimize, and deploy AI and HPC applications on AMD Instinct GPUs using the ROCm and HIP.

• What you’ll demonstrate:
  • ROCm fundamentals and driver/runtime configuration
  • HIP programming basics and porting concepts
  • Optimizing kernels and memory movement for AMD GPUs
  • Building, running, and profiling AI/HPC workloads on a single GPU
• Who it’s for:
  • Developers new to AMD GPUs
  • Practitioners moving workloads from other GPU platforms
  • Students and researchers seeking an industry-recognized credential
• Outcomes:
  • Confidence deploying on the AMD Instinct GPUs
  • A recognized signal to employers in a rapidly growing AI/HPC market

Level 2: Professional (launching later this year) Scale your AMD GPU expertise from a single device to production multi-GPU systems. This certification validates your ability to build and optimize distributed AI training and high-throughput inference on the AMD GPUs.

• Core skills:
  • Multi-GPU training with RCCL and data/tensor parallel patterns
  • Real-time inference optimization: KV-cache management, speculative decoding, continuous batching
  • Serving at scale with leading frameworks such as vLLM and SGLang
  • Multi-node distributed computing patterns and performance tuning
  • Emerging edge/robotics deployment strategies
• Who it’s for:
  • ML engineers and platform engineers taking models from lab to production
  • Teams that need predictable throughput and latency under SLA
• Outcomes:
  • The ability to design and operate distributed training and inference stacks on AMD GPUs

Level 3: Expert (Launching early 2027) Prove you can architect, deploy, and operate production AMD GPU infrastructure end-to-end. Assessment centers on a capstone project that mirrors real-world constraints and success criteria.

• What’s assessed:
  • Optimizing real AI workloads across multi-GPU clusters
  • Building production MLOps pipelines on Kubernetes with AMD AI Studio
  • Managing fleet-scale GPU infrastructure with monitoring, observability, and lifecycle automation
• Who it’s for:
  • Senior ML/AI platform engineers and architects
  • Teams standardizing on AMD Instinct for cost/performance and scale
• Outcomes:
  • Demonstrated capability to take AMD Instinct deployments from proof-of-concept to production and keep them running efficiently

What to Expect Throughout the Certification Journey

• Hands-on labs and practical scenarios that mirror real cluster environments
• Performance-focused tasks: profiling, bottleneck diagnosis, and optimization
• Toolchain depth: ROCm, HIP, RCCL, vLLM, SGLang, Kubernetes, AMD AI Studio, and supporting observability/CI/CD tooling
• Clear rubrics and artifacts you can showcase to employers

Preparing for Certification

For the Associate level, brush up on the ROCm fundamentals, HIP programming, and basic GPU performance concepts (memory hierarchy, occupancy, streams). For the Professional certification, gain experience with distributed training using RCCL, deploy inference workloads with vLLM or SGLang, and practice measuring throughput, latency, and scaling efficiency across multiple GPUs. And if you are aiming for the Expert level, build a small Kubernetes-based GPU environment, integrate AMD AI Studio, add monitoring/alerting, and run an end-to-end model lifecycle - from training and packaging through deployment, observability, and continuous optimization.

Start Your ROCm Certification Journey

Whether you're beginning your journey with AMD accelerated computing or building production-scale AI infrastructure, the ROCm Certification Program provides a clear, hands-on path to develop and validate the skills that matter. Each certification level builds on the last, helping you demonstrate real-world expertise with the AMD software ecosystem while advancing your career in AI and HPC.

To learn more about the ROCm Certification Program, visit developer.amd.com.

Heading to Advancing AI 2026? Make the most of your experience by enrolling in an in-person ROCm Certification course. Learn directly from AMD experts, complete hands-on labs, and earn your ROCm Certification badge while connecting with fellow AI developers, researchers, and platform engineers.