In May I said I'd post an update the moment the Android app cleared review. It did, and a lot more has happened since. The repo just crossed 1,000 scenario questions.
If you missed the first post, Your Next DevOps Interview Just Got Easier introduced a GitHub repo of scenario-based DevOps interview questions. They're written the way interviewers really ask them. Not "what is a pod," but "your Kubernetes nodes are NotReady at 2am, what do you check first?"
At the time it had 770+ scenarios across 11 domains and a website. The Android app was still in review.
Here's where it stands now: 1,000 questions across 12 domains, a website, and a live offline app. This post covers what changed, where it's weak, and how you can add to it. It's open source under the MIT license.
What's in the repo
The repo holds 1,000 scenario questions across 12 domains. Each domain has its own folder with one scenarios.md file.
Every question has a level tag. L1 is a fresher with 0 to 1 years of experience. L2 is mid-level, 2 to 4 years. L3 is senior or SRE, 5+ years. Each one also has a short note on what the interviewer is really testing, then a plain-language answer with steps and commands.
The levels are judgment calls. If one looks wrong to you, say so.
Git and AI engineering
Two domains deserve a note. Git was already in the repo in May, and it reached the Android app in this update. AI engineering is new everywhere.
Git is 50 questions on branching, rebasing, conflicts, recovery and history rewriting. Git looks small until someone asks you to get back something you just destroyed.
AI engineering is 70 questions on LLM foundations, RAG, fine-tuning, inference, agents and LLMOps. Why does that belong in a DevOps repo? Platform and cloud engineers are increasingly the ones asked to run model servers, keep GPU costs under control and work out why a chatbot started giving wrong answers. Those are operations problems.
This section uses a concept-first format. Each module opens with plain-English questions. It explains an idea with an everyday analogy before using jargon, and it defines every acronym on first use. Then it moves into production scenarios.
It also holds one firm rule: separate retrieval failures from generation failures. When a RAG system answers badly, the right document often never came back. That's a retrieval bug, not a model bug, and you fix it in a different place.
Read it anywhere
The same 1,000 questions live in three places. Use whichever fits the moment. The repo is where the content lives. Read it on GitHub, or clone it and grep for a topic.
The website has instant search across every question, filters for L1, L2 and L3, and dark and light themes. No ads, no tracking.
The Android app works fully offline once it's installed. You can bookmark questions and copy any question and answer to share or revise. There are no ads and no sign-up, and the Play listing says it collects no data. This update also improved search: matching domains show up as shortcuts, whole-word matches rank first, and the old 50-result limit is gone.
How to contribute
You don't need to know everything. You need to know one thing well and be willing to write it down plainly. This section follows the repo's CONTRIBUTING.md, so read that too.
Four kinds of contribution help most. New scenarios, especially your own on-call incidents with sensitive details removed. Better answers to existing questions. Proofreading: typos, broken links and outdated tool advice. And diagrams, in ASCII or Mermaid. If you have scenarios on a topic the repo doesn't cover, open an issue first to propose a new domain.
Some gaps are easy to name. In May I said Observability, Networking and Security needed the most help. They still have fewer questions than the big domains: 80, 50 and 40, against 100 or more for Kubernetes, AWS, CI/CD, Terraform, Docker and Linux/SRE. If you've lived through an incident in one of those areas, that's where a single scenario counts the most.
Search the repo for your topic before you write, so you don't duplicate. Then follow the format
`Q. [L]
What the interviewer is testing:
Answer:
---`
A good scenario describes a real situation, not just "define X." The answer runs two to five paragraphs unless the problem needs more. Include steps and commands where they apply, and explain the why, not only the what. If there are several valid approaches, say so. Avoid strong opinions without reasoning: "for this use case, Docker is often preferred because..." beats "Docker is better than Podman."
AI engineering questions have a few extra rules. Use an everyday analogy before the term. Define every acronym on first use. Separate retrieval failures from generation failures. And write original explanations instead of copying from other sites.
Each domain has its own folder, named for the domain: kubernetes, aws, ci-cd, terraform, docker, linux-sre, observability, networking, security, git, ai-engineering and general-devops. Pick the closest match, and use general-devops when nothing fits.
What's next
Next up is a new AI inference engineering section, with a playground.
Inference is where AI meets operations. It's about how a model actually serves requests, and why one setup is fast and cheap while another is slow and expensive. Reading that a setting changes latency is one thing. Seeing it change is another. The goal is to let you try a concept instead of only reading about it, so new ideas stick faster.
I won't promise a date. I'd rather ship it when it works. If there's an inference topic you always found confusing, tell me in the comments. It might become one of the first things the playground explains.
Over to you
What's the strangest production incident you've debugged? Drop a one-line version in the comments. The good ones make great interview scenarios, and you can turn yours into a pull request.
If the repo helps you, there are three other small things you can do. Star it so others can find it. Send it to someone with an interview coming up. Open an issue when you find an answer that's wrong or out of date.
A thousand questions was the easy milestone. Making them right is the harder one, and that's not a job for one person.
Find everything
GitHub repo: https://github.com/Techikrish/devops-cloud-interview-scenarios
Website: https://techikrish.github.io/devops-cloud-interview-scenarios/
Android app: https://play.google.com/store/apps/details?id=com.techikrish.devopsinterviewhub
How to contribute: CONTRIBUTING.md
Found a wrong answer: open an issue


Top comments (0)