“I’m not from AI… can I even do this?”
When I first decided to prepare for the NVIDIA Certified Professional – AI Infrastructure, I had a very honest thought:
“I’m not from a pure AI background… how am I going to understand all this?”
Yes, I did have some exposure to AI/ML during my PhD.
But here’s the truth:
That didn’t automatically mean I was ready for this exam.
Because this certification is not about:
- Building models
- Training algorithms
It’s about something very different:
AI Infrastructure
And that requires a completely different way of thinking. I started the course thinking:
“Okay, let’s just go through the modules.”
But within a few days, I was completely lost. One moment I was learning about:
- GPUs and CUDA
- Then suddenly NIM and Triton
- Then networking concepts like RDMA
- Then tools like Run:ai
It felt like:
- Everything is important
- But nothing is connected
And honestly… it was frustrating.
The moment everything changed
One day, I stopped studying and asked myself:
“What is AI actually made of in the real world?”
Not just models. Not just algorithms.
But the system behind it. And that’s when it clicked:
- AI is not just intelligence.
- AI is infrastructure.
Suddenly, my background became an advantage
This was the biggest shift. Instead of feeling like I was behind. I realized I already understood a big part of the puzzle.
Because AI infrastructure is built on:
- Compute
- Networking
- Storage
- Orchestration
Things I had already worked with. So I wasn’t starting from zero. I just needed to connect the dots differently.
So I rebuilt my learning approach
Instead of going topic by topic randomly, I created structure. I divided everything into 6 core domains:
Fundamentals — “Why AI needs GPUs”
- GPU vs CPU
- Parallel computing
- AI Factory
This is where confusion starts to clear.
NVIDIA AI Stack — “How AI actually runs”
- NGC
- NIM
- Triton
- AI Enterprise
These are not just tools — they are the runtime of AI.
Compute — “The power layer”
- A100, H100 GPUs
- Grace CPU
- Data center design
Hardware becomes meaningful when seen through AI workloads.
Infrastructure — “Where everything comes together”
- Run:ai → GPU scheduling
- Kubernetes Operators → Automation
- Mission Control → Monitoring
This is real-world AI operations.
Networking — “The part I underestimated”
- RDMA
- InfiniBand
- BlueField
Networking = performance in AI.
Data Science — “Important, but not everything”
- RAPIDS
Useful, but not the core focus here.
The biggest mistake I was making
My notes.
They looked like:
❌ Long paragraphs
❌ Hard to revise
❌ Easy to forget
The simple system that fixed everything
I switched to a structured approach:
- What is it?
- How it is correlated with Infrastructure components?
- Why does it matter?
- Key components
- Real-world use case
- Exam insights
And instead of writing long explanations, I wrote like this:
GPU:
- Parallel processing
- Used in training
- Faster than CPU
Simple. Clear. Effective.
What I learned from this journey
This wasn’t just exam preparation. It changed how I see AI completely.
- AI is not just models — it’s systems
- Prior AI/ML experience ≠ readiness for AI infrastructure
- Infrastructure engineers have a huge role in AI
- Networking is as critical as compute
- Kubernetes + GPUs = the future stack
If you’re coming from a non-AI or infra background
Let me say this clearly:
- You are not behind.
- You are just looking at AI from a different angle.
And that angle is extremely valuable.
This journey started with doubt. Even with prior exposure to AI/ML, I realized:
- Understanding models is one thing
- Understanding how they run at scale is another
Today, I feel More structured, more confident, and more aligned with where AI is heading.
What’s next?
In my next blog, I’ll break down:
GPU vs CPU: Why GPUs Power Modern AI (from an infrastructure perspective)
If you’re transitioning into AI from compute/infra, or even from an AI/ML background — let’s connect !
About The Author
