Cloud Horizon AI / Use cases

The LLM API for Microsoft Certified Trainers.

Build practice exams, translate course material, auto-grade labs, draft hands-on guides. Student submissions and your course content stay in the EU. Designed by people who teach Azure to European cohorts.

Join the waitlist Read the docs

Six workflows that buy back your weekends

The slow parts of running a training practice. Each one a few hours a week, every week.

Practice exam generator

Pain
You teach AZ-104 to twelve students every quarter. Microsoft refreshes the question bank twice a year. You either pay for a third-party item bank or write fifty new questions by hand.

Fix
Feed the published exam objectives to the API, get back fifty multiple-choice items with correct-answer rationale and distractor logic. Spot-check, edit the ones the model missed, ship.

AZ-104, AZ-204, AZ-305MS-102, MS-700, MS-900AI-102, DP-203, SC-200

Course material translation

Pain
Your students are Dutch, German, and French. Microsoft Learn is mostly English. Translating slide decks by hand kills your prep budget.

Fix
Pipe markdown decks through MiniMax M2.5 with a glossary of preserved technical terms. Keep code blocks intact, translate prose. Ship the same lab in seven languages from one source.

DE, FR, NL, IT, ES, PL, SVPreserve PowerShell, Bicep, KQLGlossary-aware terminology

Lab walkthrough auto-grading

Pain
Twelve students submit lab evidence. You spend Saturday reading screenshots and writing feedback.

Fix
Students submit a markdown report. Model scores against your rubric and drafts personalized feedback. You review the borderline ones, sign off in a third of the time.

Az CLI command sequencesBicep deployment outputsNetwork ACL configurations

Tailored explanations on the fly

Pain
Mixed-experience cohort. The senior infra engineer is bored, the junior helpdesk hire is lost.

Fix
Two prompts, two explanation depths. Same source content, different audience framing. Drop both into the LMS, students pick.

Beginner / advanced toggleWorked-example expansionsAnalogy generation

Hands-on lab guide drafting

Pain
Microsoft Learn labs do not always match your tenant config. You rewrite the steps every cohort.

Fix
Paste the official lab steps and your tenant constraints. Model returns adjusted steps with the right resource group naming and SKU substitutions.

Tenant-aware path rewritesSKU availability swapsRegion-specific tweaks

Office hours triage

Pain
Same five questions every cohort. You answer them in DM, by email, in the Teams channel.

Fix
Ingest your past answers, run a small RAG layer with embeddings. Students get a first-pass answer in chat, you handle the genuinely new ones.

Embedded FAQ over course contentSlack/Teams bot integrationEscalation when the model is unsure

Recommended model per exam track

Six exam tracks we have tested against. The model column is what we would pick for question generation and lab drafting on that track.

Exam	Title	Model	Why this one
AZ-104	Microsoft Azure Administrator	kimi-k2.5	Long-context syllabus reasoning
AZ-305	Designing Microsoft Azure Infrastructure Solutions	glm-4.6	Architecture trade-off rationale
AI-102	Designing and Implementing a Microsoft Azure AI Solution	qwen-3-coder	Code-heavy lab generation
MS-102	Microsoft 365 Administrator	minimax-m2.5	Multilingual scenarios
SC-200	Microsoft Security Operations Analyst	glm-4.6	KQL and alert tuning
DP-203	Data Engineering on Microsoft Azure	kimi-k2.5	Long pipelines, multi-step reasoning

One curl, twelve practice items

Strict JSON output mode plus a system prompt your reviewers wrote once. Same shape every cohort, no parsing surprises.

# Generate twelve AZ-104 practice items from the Microsoft Learn objectives
curl https://api.cloudhorizons.ai/v1/chat/completions \
  -H "Authorization: Bearer $CLOUD_HORIZONS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k2.5",
    "messages": [
      {"role": "system", "content": "You write MCT-grade practice questions. Each item: stem, four options, one correct, rationale per option."},
      {"role": "user", "content": "Topic: Manage Azure identities and governance. Twelve items, mixed difficulty."}
    ],
    "response_format": {"type": "json_object"}
  }'

Why MCTs pick us

EU student data residency without a procurement battle

Universities and large corporates increasingly require EU residency for any tool that sees student work. We are operated under Spot Cloud B.V. (KvK 89708873), governed by Dutch law, with inference in Amsterdam and Frankfurt. Standard EU Data Processing Agreement on team plans. No US parent in the chain.

Join the waitlist