- Vishakha Sadhwani
- Posts
- Bad, Good, Excellent: The Learning Ladder for Cloud, DevOps and AI
Bad, Good, Excellent: The Learning Ladder for Cloud, DevOps and AI
Nine skills, three ways to learn each, and the free resources to climb from watching to building.
Vishakha Sadhwani
October 07, 2026
Every skill in tech can be learned three ways.
Bad is consuming. You watch, read, and highlight, and it feels productive, but nothing sticks. Good is guided doing. Someone sets up the environment, you follow the steps, and you start to build muscle memory. Excellent is building or breaking things on your own. There's no script, things fail, and you work out why.
You can't skip straight to Excellent, but you shouldn't stay on Good forever either. Use Good to get unstuck, then climb.
Below are nine skills, each with the ladder, the free resources to climb it, and the crash courses that fit. Bad options get no links on purpose.
1. Projects
Bad: Tech tutorials
Good: Follow-along projects → practical-tutorials/project-based-learning
Excellent: Build Your Own X (OS, browser, database, shell, Git) → codecrafters-io/build-your-own-x
Why the jump matters: A tutorial shows you how to use a tool. Rebuilding a small version of Git or a shell shows you how it works, and that's what interviewers probe.
Free resources
End-to-End DevOps + AIOps Project (complete series): a follow-along project that ties many tools together
Docker CI/CD project: push, build, ship
2. Cloud
Bad: Reading AWS/GCP/Azure docs end to end, or binge-watching Cloud Digital Leader videos
Good: AWS Cloud Quest, a role-playing game where you solve cloud tasks in a real AWS console (the Cloud Practitioner role is free)
Excellent: AWS Free Tier → Build & Deploy
Why the jump matters: Cloud Quest guides you. Your own account doesn't. You pick the services, wire them together, debug the permissions, and pay for your mistakes, which is why you remember them.
Free resources
The Cloud Resume Challenge: a classic first "build and deploy" project
Terraform with AWS: Real-Time Project · companion repo: VPC, subnets, load balancer and instances, all as code
Crash courses
Before you build: Set a budget alert first, and tear down everything when you're done. Check the Free Tier page for what new accounts currently get, because the offer has changed over time.
3. Cloud Labs
Bad: YouTube tutorials
Good: Google Cloud Skills Boost, hands-on labs in a real, temporary Google Cloud project. Join the free Google Cloud Innovators program for 35 free lab credits every month.
Excellent: escbash → Rebuild
Why the jump matters: escbash gives you a real Linux machine and grades the state of the machine, not your clicks. The "Rebuild" step is where you learn the most: after the lab is graded, rebuild the same thing from memory in your own account.
What's free on escbash: The Linux and Python skills are free in full, and the first topic of every other skill is free. The DevOps, Cloud and AI tracks are paid.
4. Networking
Bad: Networking theory, or reading AWS networking docs
Good: Cisco Packet Tracer, a free network simulator. You get the download by enrolling in the free Getting Started with Cisco Packet Tracer course.
Excellent: AWS Networking Immersion Day labs → Troubleshoot
Why the jump matters: Packet Tracer teaches you how packets move between routers and switches. AWS VPC labs teach you why your EC2 instance can't reach the internet. The real skill is the second one, so once a lab works, break a route table or security group on purpose and fix it.
Cisco Packet Tracer: follow-along videos
Jeremy's IT Lab: Free CCNA 200-301 Complete Course: the most recommended free CCNA course. Most lectures have a matching Packet Tracer lab video.
Free CCNA 200-301 course playlist: starts with "Build a network with me for free using Cisco Packet Tracer"
Cisco Packet Tracer: GitHub repos
asher-lab/Cisco-Packet-Tracer-Practicals: ready-made .pkt files for DNS, DHCP, HTTP servers, Telnet, and WLAN. Open them and explore.
c4geeks/ccna-labs: 19 CCNA labs grouped by exam domain, with tested solution configs so you can attempt first and check after (walkthrough guide)
GibJaf/CCNAv7: solved Packet Tracer activities from the Cisco CCNAv7 course
AWS VPC troubleshooting
VPC flow logs: concept · step by step. Flow logs are how you see what got rejected and why.
Crash course
Cost note: The AWS networking labs use NAT gateways and Transit Gateways, which aren't free. Finish in one sitting, then delete everything.
5. Linux
Bad: Linux cheat sheets
Good: Linux Journey. The old linuxjourney.com now redirects here, and it's still free.
Excellent: OverTheWire Bandit
Why the jump matters: Linux Journey explains the commands. Bandit gives you a password hidden somewhere on a server and makes you find it, so you have to combine commands to solve a problem.
Free resources
escbash Linux skill: free in full, graded on a real VM
SadServers: broken Linux servers to fix, a good next step after Bandit
6. Kubernetes
Bad: Kubernetes videos on their own
Good: Kubernetes Basics (official tutorial): deploy, expose, scale and update your first app
Excellent: Killercoda → Deploy & Troubleshoot
Why the jump matters: Getting a pod running is easy. Working out why it's stuck in CrashLoopBackOff is the job. Killercoda gives you free in-browser clusters with broken scenarios to fix.
Free resources
Killercoda playgrounds: empty clusters for your own experiments
Crash courses
7. Certifications
Bad: Random certifications
Good: AWS Cloud Practitioner / AZ-900 / Cloud Digital Leader
Excellent: CKA / CKAD / CKS
Why the jump matters: The foundational exams are multiple choice. They prove you know the vocabulary, which helps you get past HR filters. The Kubernetes exams are hands-on: you get a live terminal and two hours, so they prove you can do the work.
Good: Foundational certifications
Cert | Exam page (register here) | Free prep |
|---|---|---|
AWS Certified Cloud Practitioner (CLF-C02) | AWS Skill Builder: search "AWS Cloud Practitioner Essentials" · AWS Cloud Quest (Cloud Practitioner role) | |
Microsoft Azure Fundamentals (AZ-900) | learn.microsoft.com/credentials/certifications/azure-fundamentals | |
Google Cloud Digital Leader | Google Cloud Skills Boost: search "Cloud Digital Leader" learning path |
Each exam page has the official exam guide and a Schedule / Register button.
Excellent: Kubernetes certifications
Cert | Exam page (register here) | Free prep |
|---|---|---|
CKA (Administrator) | ||
CKAD (Developer) | ||
CKS (Security) |
Free foundation for all three: Introduction to Kubernetes (LFS158) from the Linux Foundation.
Good to know: A Linux Foundation exam registration includes one free retake and two sessions on the killer.sh exam simulator. CKS requires an active CKA. The Linux Foundation runs frequent sales (Cyber Monday is the big one), so don't pay full price unless you're in a hurry.
8. System Design
Bad: Reading finished system design solutions
Good: Hand-drawing architectures against the AWS Well-Architected Framework using Excalidraw, a free whiteboard
Excellent: Whiteboard system design → Mock interview
Why the jump matters: Reading a solution tells you the answer. Drawing it yourself shows you what you don't know. A mock interview shows you whether you can explain your design while someone pushes back.
Free resources
donnemartin/system-design-primer: the standard free reference
Exponent peer mock interviews (formerly Pramp): free accounts get a few peer mock credits a month. You interview a peer, then they interview you.
Crash courses
9. AI Tools
Bad: Learning tools in isolation (one LangChain tutorial, one MLflow video, one Docker course)
Good: Connecting tools through many projects → Made With ML · MLOps-Basics
Excellent: Building an end-to-end deployment pipeline: Data → training (MLflow) → CI/CD (GitHub Actions) → Docker → Kubernetes → monitoring (Evidently) → MLOps Zoomcamp · LLM Zoomcamp
Why the jump matters: Anyone can call an API in a notebook. Running a model that retrains, redeploys and alerts you when it drifts is what companies actually pay for.
Follow-along videos
Learn MLOps by Creating a YouTube Sentiment Analyzer (freeCodeCamp, ~3 hrs): MLflow, DVC, Docker, AWS and Flask in one project
Follow-along repos
JIAZHEN/machine-learning-mlops: MLflow, FastAPI, Evidently, Docker and Kubernetes. CI/CD isn't built yet, so adding the GitHub Actions pipeline is your "Excellent" step.
aashu-0/MLOps_Learning_Project: full CI/CD to AWS EKS
Crash courses
Crash Course Quick List
Networking: https://youtu.be/bEFAFHIahXk
VPC (hands-on): https://www.youtube.com/watch?v=N6qtN0c_e9A
VPC flow logs: https://www.youtube.com/watch?v=j4ab0R_XPTA · https://www.youtube.com/watch?v=2PQIDssp9ts
AWS / GCP / Azure roadmaps: https://youtu.be/Kuy-pGuz02M · https://youtu.be/CTIJWijru9E · https://youtu.be/liRgZeF6mbk
System design: https://youtu.be/ihQQJuKHY7A
CI/CD: https://youtu.be/ixNNyLcWXX8
Docker CI/CD project: https://youtu.be/M2fOJA6U5PE
End-to-end DevOps + AIOps project: https://www.youtube.com/playlist?list=PLXkUFcIv0_b7rzZe0o_2-GOS2qn5-0OQy
How to Use This
Don't try to reach Excellent on all nine skills at once. Pick one, spend a week on Good, then move to Excellent as soon as Good starts to feel easy. Feeling comfortable is the signal that you've stopped learning.
The thread running through all of this is the same: stop consuming, start breaking things. Break the route table. Kill the pod. Delete the config and rebuild it from memory. That's the part nobody can watch for you.
Free tiers, credits, exam prices and course availability change often. Check the official page before you pay for or deploy anything, and search the title if a link has moved.