supraj.dev

Free · No sign-up · Model answers

Interview Studio

Real interview questions for DevOps, cloud and AI engineers. Answer out loud, open the model answer, then rate yourself.

Questions
1,718
Topics
35
Difficulty mix
107 easy791 medium820 hard

Pick your path

What role are you preparing for?

1,208 questions · 18 topics

DevOps Engineer

CI/CD, containers, infrastructure as code and the incidents that test them.

  • Kubernetes
  • Docker
  • Terraform
  • Linux
  • AWS Architect
  • Git
  • +12
Show DevOps topics

890 questions · 7 topics

Cloud Engineer

Architecture, networking, IAM and resilience on AWS and Google Cloud.

  • AWS Architect
  • GCP
  • Terraform
  • Kubernetes
  • Docker
  • Linux
  • +1
Show Cloud topics

510 questions · 17 topics

AI / ML Engineer

LLMs, RAG, agents, training, serving and the cost of running them.

  • LLM fundamentals
  • Prompt & Context Engineering
  • RAG
  • Vector Search
  • Agents, tools & memory
  • MCP & Tool Calling
  • +11
Show AI topics

1,238 questions · 19 topics

Platform / SRE

Reliability, observability, service mesh and production debugging.

  • Kubernetes
  • Prometheus
  • Linux
  • AWS Architect
  • Istio
  • Terraform
  • +13
Show Platform topics

All topics

Choose a topic, start anywhere.

35 topics · 1718 questions

AWS Architect

178 questions

Solutions architecture, IAM, VPC, compute, and resilience.

14 easy, 81 medium, 83 hard.

Kubernetes

164 questions

Pods, scheduling, networking, and the control plane.

8 easy, 68 medium, 88 hard.

Linux

148 questions

Processes, signals, networking, and troubleshooting.

5 easy, 63 medium, 80 hard.

Docker

141 questions

Images, layers, networking, and container internals.

5 easy, 64 medium, 72 hard.

Terraform

136 questions

State, plan/apply, modules, and drift.

4 easy, 51 medium, 81 hard.

GCP

110 questions

Google Cloud architecture, networking, IAM, and SRE operations.

1 easy, 46 medium, 63 hard.

Git

70 questions

Branching, merging, rebasing, and recovery.

11 easy, 31 medium, 28 hard.

Production evaluation, observability & cost

59 questions

Measure useful outcomes, trace failures and control the cost of successful work.

1 easy, 33 medium, 25 hard.

Helm

53 questions

Charts, releases, hooks, and lifecycle management.

7 easy, 27 medium, 19 hard.

Harness

48 questions

Pipelines, CV, secrets, and multi-cloud deployments.

8 easy, 19 medium, 21 hard.

LLM training, fine-tuning & mathematics

45 questions

Pretraining objectives, LoRA and QLoRA, distillation, losses, gradients, PCA and adaptation quality.

1 easy, 21 medium, 23 hard.

AI platform serving & scaling

43 questions

Find the serving bottleneck before adding GPUs. Balance latency, throughput and memory.

2 easy, 23 medium, 18 hard.

Agents, tools & memory

38 questions

Bound agent loops, connect MCP tools and recover state without repeating side effects.

2 easy, 19 medium, 17 hard.

RAG

37 questions

Retrieval, grounding and citations: follow the evidence through the pipeline.

3 easy, 23 medium, 11 hard.

CUDA

33 questions

GPU threads, streams and attention kernels explained through practical questions.

2 easy, 19 medium, 12 hard.

Agent security & sandboxing

30 questions

Keep untrusted instructions away from authority. Bound tools, credentials and execution.

1 easy, 11 medium, 18 hard.

GitHub Actions

30 questions

Workflows, actions, runners, and CI/CD automation.

0 easy, 18 medium, 12 hard.

LLM Evals & Observability

30 questions

Golden sets, judges, CI gates and traces that explain a bad answer.

1 easy, 15 medium, 14 hard.

LLM fundamentals

29 questions

Tokens, context, sampling and output contracts. Explain what a model can and cannot guarantee.

12 easy, 15 medium, 2 hard.

Vector Search

29 questions

Embeddings, ANN indexes, filters, hybrid ranking and reindexing at scale.

1 easy, 14 medium, 14 hard.

Transformers, attention & architecture

27 questions

Q/K/V attention, encoders and decoders, positional signals, MoE, embeddings and model design maths.

1 easy, 16 medium, 10 hard.

MCP & Tool Calling

27 questions

Tool contracts, authorization and safe side effects for agents that act.

1 easy, 12 medium, 14 hard.

AI system design

25 questions

End-to-end designs for LLM platforms: requirements, architecture, failure modes, observability and cost.

0 easy, 6 medium, 19 hard.

PyTorch

25 questions

Tensor memory, strides, views and the hidden cost of layout changes.

1 easy, 13 medium, 11 hard.

Go

24 questions

Concurrency, memory model, runtime, and SRE tooling.

1 easy, 10 medium, 13 hard.

Python

23 questions

Concurrency, performance, async, and SRE tooling.

1 easy, 11 medium, 11 hard.

Prometheus

21 questions

Metrics collection, PromQL, alerting, and observability.

2 easy, 12 medium, 7 hard.

GPU economics & capacity planning

19 questions

Commitments, reservations, spot, utilisation, power and chargeback: GPU capacity that finance can sign off.

0 easy, 12 medium, 7 hard.

ArgoCD

14 questions

GitOps, ApplicationSets, sync strategies, and multi-cluster.

2 easy, 7 medium, 5 hard.

Istio

14 questions

Service mesh, traffic management, security, and observability.

2 easy, 7 medium, 5 hard.

Ansible

13 questions

Playbooks, roles, automation, and configuration management.

3 easy, 6 medium, 4 hard.

HashiCorp Vault

13 questions

Secrets management, encryption, PKI, and access control.

2 easy, 6 medium, 5 hard.

Jenkins

8 questions

CI/CD triggers, reproducible builds, failure diagnosis and safe recovery.

0 easy, 5 medium, 3 hard.

AI on Kubernetes

7 questions

GPU scheduling, model loading, queueing and inference routing on clusters.

1 easy, 2 medium, 4 hard.

Prompt & Context Engineering

7 questions

System prompts, structured output, caching and what goes into the window.

1 easy, 5 medium, 1 hard.

Watch it explained

Short videos for the hard parts.

How to practise here

  1. Pick a topicFilter by easy, medium or hard, or search for the term your interviewer used.
  2. Answer out loudSay your answer before you open the model answer. That is the part that sticks.
  3. Rate yourselfKnew it, Partly or Missed. Your progress stays in this browser, no account needed.
  4. Practice modeOne question at a time, with keyboard shortcuts: Space to reveal, 1–3 to rate.