From delivered hardware to live inference.
Hands-on GPU cluster bring-up training. On-site or remote, for the teams that build and run AI datacentres.
GPU cluster training
Contact for a quote
Practitioner-led. Shaaz Ahmed — Senior Principal Engineer with GPU datacentre bring-up experience at HUMAIN.
The engagement
Bring up the cluster.
Build the team.
Delivered servers are not a working cluster. Burn-in failures, misconfigured networking and unfamiliar software stacks can leave expensive hardware idle.
This is practical training, delivered on-site or remotely, to take your team from server bring-up to production inference—not just through a slide deck.
GPU cluster training / Included
- Verify the physical foundation
- Rack, power and cabling checks. Server bring-up, including Dell AMD MI355X-class nodes.
- Validate healthy nodes
- Firmware, drivers and ROCm/CUDA stacks. Burn-in and health validation before moving up the stack.
- Make the fabric work
- Cluster networking, including Cisco fabrics and RoCE. Topology checks and performance testing.
- Operate and troubleshoot
- Scheduling, orchestration, observability and failure triage. Learn to diagnose the cluster, not just deploy it.
- Serve real inference
- Get LLM inference serving live and tuned on the new cluster.
Who it is for / Where it happens
For the team that has to make it run.
Datacentre operators, neoclouds and GPU cloud providers. Enterprise infrastructure teams standing up their first GPU clusters.
Sovereign AI and government programmes, including teams in the Gulf and Europe. Training is available on-site or remotely; the engagement starts with your infrastructure and goals.
Your instructor
Shaaz Ahmed
Founder, AI Datacenter Labs
Senior Principal Engineer
Taught by a practitioner who has brought up GPU datacentres at HUMAIN, Saudi Arabia’s sovereign AI company.
Before HUMAIN, Shaaz built distributed systems and worked on formal verification at AWS Lambda for approximately five years, and was a staff engineer on Poe.com at Quora.
MSc in HPCTrinity College Dublin
MSc in Software EngineeringUniversity of Oxford
How an engagement works
Start with your cluster.
-
01 / SCOPE
Scoping call
Discuss your hardware, team and bring-up goals.
-
02 / PLAN
Tailored training plan
Shape the curriculum around the work your team needs to do.
-
03 / PRACTISE
Hands-on training
Work through bring-up, validation, networking and inference.
-
04 / HAND OVER
Handover to your team
Connect the training to ongoing operation and failure triage.
Before we talk.
Can training happen remotely?
Yes. Practical training is available on-site or remotely. Use the scoping conversation to discuss which delivery format suits your team and infrastructure.
Which hardware and software does it cover?
The curriculum includes Dell AMD MI355X-class nodes, ROCm/CUDA stacks, Cisco fabrics and RoCE. Bring your actual hardware and software stack to the scoping call so the plan can be tailored.
Is this just server installation?
No. The training continues through health validation, cluster networking, scheduling, observability and failure triage to getting LLM inference live and tuned.
How is an engagement priced?
Training is quoted per engagement or per team. Request a quote with your cluster context, training goals and preferred delivery format.
Put your hardware to work.
GPU cluster training · On-site or remote
Request a quote