Cloud skills teams need (plus 6 hands-on learning projects)

There’s a pattern I’m seeing again and again right now.

Leaders are saying, “We need AI.” So, teams are experimenting, and pilots are everywhere. And yet very little is actually making it into production.

That’s not a tooling problem. It’s not even really an AI problem. It’s a cloud skills problem.

Cloud isn’t optional anymore; it’s the foundation. After all, AI doesn’t exist in isolation. Every AI system depends on scalable infrastructure, secure systems, and reliable data pipelines. In other words, it all relies on cloud.

Without that foundation, AI initiatives fail—not because the model wasn’t good enough, but because the environment wasn’t ready. So, what cloud skills do teams actually need?

Understanding architecture, networking, compute, storage, and the shared responsibility model is essential. This is the baseline that everything else builds on.

That means knowing how the major platforms (AWS, Azure, and Google Cloud) are structured, how virtual networks and subnets work, and how compute and storage options differ depending on the workload. It also means understanding what the cloud provider is responsible for securing versus what your team owns.

Ask team members to spin up a simple multi-tier application on their cloud platform of choice. The goal is exposure, not a perfect build. Getting hands-on with VPCs, IAM roles, and storage buckets in a real environment builds intuition that no amount of reading can replicate. Cloud provider sandboxes and labs are a low-risk way to do this.

Teams need to design scalable systems and align technology decisions to business outcomes, not just adopt tools because they’re new or popular.

Concretely, that means understanding auto-scaling and load balancing so systems can handle variable demand without over-provisioning. It also means knowing when to use microservices versus a simpler architecture, and how to design for fault tolerance using patterns like multi-region deployments or availability zones. 

On the soft skills side, it means being able to translate a business requirement into an infrastructure decision, and push back when a shiny new tool doesn't actually solve the problem.

Run architecture review sessions where team members present a proposed design and the group stress tests it. Ask questions like: 

• What happens when this component fails? 

• How does this scale at 10x the current load? 

• What does this cost at scale? 

This builds both technical and communication muscles at the same time.

Using tools like Terraform or CloudFormation allows teams to create repeatable, reliable environments and reduce risk as systems scale. AWS CDK and Pulumi are also worth knowing, particularly for teams who prefer working in general-purpose programming languages rather than domain-specific ones.

The skill here isn't just knowing the syntax. It's understanding how to structure code for reuse, manage state safely, and integrate IaC into a CI/CD pipeline so infrastructure changes go through the same review process as application code.

Give teams a real problem to solve, like provisioning a staging environment that mirrors production, and ask them to do it entirely in code. No clicking around in the console. Then have them tear it down and rebuild it. If it works the second time without manual intervention, they're on the right track.

Managing performance, cost, and scalability in production environments is critical, especially as AI workloads grow more complex.

This means understanding how to work with managed services like Amazon SageMaker, Azure Machine Learning, or Google Vertex AI, and how to build and maintain the data pipelines that feed them. Tools like Apache Airflow for orchestration, and services like AWS Glue or Azure Data Factory for data integration, are increasingly part of the picture.

It also means understanding GPU compute, model serving, and how to monitor a model's performance over time, not just at launch.

Have teams take an existing ML model, even a simple one, and focus entirely on the deployment side. How do you serve it reliably? How do you monitor it? What happens when the data drifts? 

Modeling is often the easy part. The operations are where most teams struggle.

Identity and access management, securing data and pipelines, and understanding modern threats are essential as AI increases the attack surface.

In practice, this means knowing how to apply least-privilege principles using tools like AWS IAM, Microsoft Entra ID, or Google Cloud IAM.

It also means understanding how to encrypt data at rest and in transit, how to manage secrets properly using tools like AWS Secrets Manager or HashiCorp Vault, and how to use cloud-native security tooling like AWS Security Hub or Microsoft Defender for Cloud to detect and respond to threats.

Run a security audit exercise. Give teams access to a deliberately misconfigured environment and ask them to find the vulnerabilities. What's over-permissioned? What's exposed? What's unencrypted? 

This kind of threat-modelling exercise builds the instinct to think like an attacker—which is exactly what they need to defend against bad actors using AI systems.

Cloud makes it easy to build, but also easy to overspend. Teams must understand how to optimize cost while maintaining performance.

That means knowing the difference between on-demand, reserved, and spot instances, and when to use each. It also means using tools like AWS Cost Explorer, Azure Cost Management, or Google Cloud's cost management suite to track spend, set budgets, and identify waste. 

On top of that, teams must understand performance monitoring tools like CloudWatch, Azure Monitor, or Datadog, to make smarter infrastructure decisions rather than just scaling up and hoping for the best.

Give teams a cost challenge. Present them with a workload and a budget, and ask them to optimize for cost while maintaining performance. What can be right-sized? What can be moved to spot? What's running when it doesn't need to be?

Putting a real constraint on the problem makes the learning stick in a way that theory never does.

AI is changing everything, but it is not replacing the fundamentals. If anything, it is making them more important than ever.

The teams that will get AI into production aren't necessarily the ones with the best models. They're the ones with the strongest foundations: people who understand how to build, secure, and operate systems at scale. 

If you are thinking about what skills your teams need next, don’t start with AI. Start with cloud.

Accelerate cloud transformation and unlock AI at scale with Cloud Ready, our cloud skills solution. Learn more.