THE DEV BENCH
🔷

Microsoft Azure

Azure administration & cert prep — AZ-104 decks, practice exams, and hands-on IaC/CLI/Azurite labs.

Start here

AZ-104 · Manage Identities & Governance36 cards — step 1 of 10

Then get hands-on: Azure IaC lab 7 scenarios.

AZ-104 Azure Administrator — learning pathExam-domain map, decks, practice exams, free hands-on IaC/CLI/Azurite labs, a skill map, and free-first resources — one workspace.AI-300 Machine Learning Operations Engineer — learning pathAzure MLOps/GenAIOps (Azure ML + Foundry) — five skill-area decks (99 cards), an 87-question scenario practice exam, a study routine, and free-first resources, verified against the official skills-measured guide.

Flashcard decks

10 decks · 285 cards · in study order

AZ-104 Administrator

5 decks · 186 cards
1AZ-104 · Manage Identities & GovernanceExam domain 1 (20–25%): Microsoft Entra ID, users & groups, Azure RBAC, Azure Policy, subscriptions & management groups, tags, resource locks, and cost management.362AZ-104 · Implement & Manage StorageExam domain 2 (15–20%): storage accounts & redundancy (LRS/ZRS/GRS/GZRS), blob tiers & lifecycle, access control (keys, SAS, stored access policies, Entra RBAC), storage firewalls, service vs private endpoints, encryption, Azure Files & File Sync, AzCopy & Storage Explorer.383AZ-104 · Deploy & Manage ComputeExam domain 3 (20–25%): ARM templates & Bicep (deployment modes, parameters, export, what-if), virtual machines (sizes, managed disks, encryption, availability sets vs zones, scale sets), containers (ACR, ACI, Container Apps), and Azure App Service (plans, scaling, deployment slots, TLS, networking, backup).374AZ-104 · Implement & Manage Virtual NetworkingExam domain 4 (15–20%): virtual networks & subnets, VNet peering, public/private IPs, user-defined routes, NSGs & ASGs, effective security rules, Azure Bastion, service vs private endpoints, Azure DNS & private DNS, load balancers, and Network Watcher troubleshooting.385AZ-104 · Monitor & Maintain ResourcesExam domain 5 (10–15%): Azure Monitor (metrics vs logs, Log Analytics, diagnostic settings, activity log, KQL, alerts, action groups, alert processing rules, Insights, the Azure Monitor Agent & DCRs), Network Watcher, and backup/recovery (Azure Backup, Recovery Services vs Backup vaults, backup policies, restore, Azure Site Recovery, RPO/RTO).37

AI-300 MLOps

5 decks · 99 cards
6AI-300 · ML Model Lifecycle & Operations (25-30%)The biggest area — running the model lifecycle in Azure Machine Learning: workspaces & assets, data/datastores, environments, training as jobs, pipelines & components, the model registry, endpoints (managed online + batch), deployments, and the automation (schedules, event triggers, retraining) that turns a notebook experiment into an operated system.287AI-300 · GenAIOps Infrastructure (20-25%)Standing up generative-AI systems in Microsoft Foundry (Azure AI Foundry): projects & hubs, the model catalog, deploying and fine-tuning models, prompt flow for orchestration, connections & Azure OpenAI, RAG with grounding data (Azure AI Search), agents, and deploying GenAI apps to endpoints — the operational plumbing behind an LLM app, not prompt-writing.238AI-300 · MLOps Infrastructure — IaC & CI/CD (15-20%)The platform plumbing: provisioning workspaces, compute, and endpoints as infrastructure-as-code (Bicep / ARM / Terraform / Azure CLI), building CI/CD for ML with GitHub Actions or Azure DevOps (environments, approvals, staged promotion), securing it with RBAC, managed identity, VNets and Key Vault, and organizing dev/test/prod — the DevOps discipline the 'operations engineer' title is really about.189AI-300 · GenAI Quality Assurance & Observability (10-15%)Making generative-AI output trustworthy and watchable: evaluating quality with metrics (groundedness, relevance, coherence, fluency, similarity) and AI-assisted + safety evaluators, content safety / responsible-AI filtering, monitoring deployed GenAI apps (Application Insights, token usage, latency, feedback), and catching regressions before and after release.1510AI-300 · Optimize GenAI Systems & Performance (10-15%)Tuning deployed systems for cost, latency, and quality: token/cost optimization, caching, model right-sizing, throughput (PTU vs pay-go, provisioned capacity, batching), retrieval-quality tuning for RAG, prompt optimization, and scaling endpoints — squeezing a production GenAI/ML system without a full rebuild.15

Practice exams

2 sets

Azure IaC lab

live
A real box with the Azure CLI, Bicep, Terraform and Azurite. Learn mode authors templates from scratch; Break-fix mode fixes ones that won't build or validate — all graded offline, no Azure account.

Bicep drills

5 drills
Write real Azure Bicep — the readable ARM-template DSL — parameters, resources, and outputs, graded live against a Bicep-shape parser in your browser. No Azure account needed.