AI-300 practice questions
Microsoft · AI-300 · 200 questions
Original practice questions for the Microsoft AI-300 exam (Microsoft Certified: Machine Learning Operations Engineer Associate), covering MLOps infrastructure on Azure Machine Learning, machine learning model lifecycle and operations, GenAIOps infrastructure with Microsoft Foundry, generative AI quality assurance and observability, and optimizing generative AI systems and model performance — framed in civic and municipal scenarios such as city AI pilots, public-health records, permitting chatbots, 311 assistants, transit / utilities operations, and courts / clerk workflows.
This course contains the use of artificial intelligence.
About the AI-300 exam
- Time allowed
- 2 hours
- Passing score
- 700 (scaled, 1-1000)
- Languages
- English
- Format
- Proctored exam through Pearson VUE; may include interactive components
Schedule this exam The certification this earns
Exam details published by the vendor, checked 22 October 2026. Vendors change fees and formats without notice — confirm on the vendor's own page before you book.
Practice Quizzes
Test your knowledge with standard 20-question practice sets.
Quiz 1
Quiz 2
Quiz 3
Quiz 4
Quiz 5
Quiz 6
Quiz 7
Quiz 8
Quiz 9
Quiz 10
Browse by Domain
Study specific topics at your own pace.
Design and implement an MLOps infrastructure · 38 questions
- A city IT office is launching a permits forecasting pilot and wants data scientists to share one governed Azure Machine Learning project boundary instead of scattering experiments across personal subscriptions. What should they create first?
- A county public-health analytics team must place a new Azure Machine Learning workspace so it inherits the same billing, policy, and network controls used for other HIPAA-scoped civic workloads. Where should the workspace be created?
- A water utility keeps training datasets in an Azure Blob container and wants Azure Machine Learning jobs to read that data in place. What should the MLOps engineer register?
- A transit authority stores ridership CSVs in Azure Data Lake Storage Gen2 and must expose them to Azure Machine Learning training jobs. Which datastore approach fits best?
- A housing authority needs overnight AutoML jobs for waitlist forecasting to run on scalable CPU capacity without leaving expensive compute online all day. Which compute choice best matches that need?
- A fire-district pilot needs GPU capacity for deep-learning smoke-detection experiments but only CPU for tabular permit-risk models. How should compute targets be matched?
- A city clerk's office wants data scientists to submit jobs to a shared Azure Machine Learning compute cluster but must prevent them from deleting the workspace. Which access approach is appropriate?
- A county courts IT team needs CI pipelines to submit Azure Machine Learning pipeline jobs without interactive user login. What identity pattern should they use?
- A parks department wants seasonal interns to view production scoring metrics in Azure Machine Learning without altering datastores. What should the MLOps lead configure?
- A 311 analytics unit finds Azure Machine Learning jobs failing after a storage account key rotation. What should the team update?
- A municipal open-data team needs Azure Machine Learning jobs to pull structured parcel attributes from Azure SQL for feature engineering. What should they configure?
- A school-district facilities team needs interactive notebook exploration of energy-usage models before any scheduled multi-node training. Which compute target fits that interactive work?
- A library consortium requires Azure Machine Learning workspace diagnostic logs to support municipal audit reviews. Where should those logs be sent?
- A permits office wants reproducible Azure Machine Learning training runs that always pin the same historical applications table. What asset should they create?
- A public-health lab finished cleaning lab-results data and must promote exploration v1 to a production training input without renaming folders by hand. What practice should they use?
- A transit ML team wants every ridership training job to use the same pinned scikit-learn and pandas versions. What should they define once and reuse?
- A water-quality model's production scoring fails because a preprocessing package that existed in training is missing from the scoring environment. What should the team fix?
- A city fraud-detection effort needs the same feature-engineering logic reused across several Azure Machine Learning pipelines. What asset type should they publish?
- A county assessor team validated a data-prep step and wants other departments to compose it into their own Azure Machine Learning pipelines. What should they do?
- Two city departments need to share a golden environment image across separate Azure Machine Learning workspaces without emailing Dockerfiles. Which capability should they use?
- A regional emergency-management partnership wants mutual-aid forecasting teams in different Azure Machine Learning workspaces to consume the same registered model-component catalog. What should they establish?
- A housing authority must keep PII-bearing resident files from being treated like public open-data assets in Azure Machine Learning. What discipline should they apply to data assets?
- A municipal climate team must patch an Azure Machine Learning environment after a dependency CVE while keeping builds reproducible. What should they do?
- A courts analytics group has a large ETL process that is hard to debug as one monolithic Azure Machine Learning step. How should they improve maintainability?
- A city MLOps lead has a vetted featurizer in a sandbox Azure Machine Learning registry and must make it available to production workspaces. What is the correct promotion action?
- A city platform team must deploy identical Azure Machine Learning workspaces to staging and production without repeating portal clicks. Which approach best meets that requirement?
- A county DevOps engineer must create Azure Machine Learning compute and datastores alongside a new workspace as part of an automated pipeline. Which approach best fits that goal?
- A transit authority needs Azure Machine Learning training jobs to pull code from a private civic GitHub repository with secure access. What should the MLOps team configure?
- A public-works MLOps workflow should provision a new Azure Machine Learning experiment compute cluster automatically when a deploy workflow runs. Which capability should they use?
- A health department requires that its Azure Machine Learning workspace accept traffic only from the municipal private network, not the public internet. What should the team implement?
- A city ML project must keep training scripts, pipeline YAML, and environment files under version control with branch protection before production deploys. Which practice should the team adopt?
- A parks analytics team must review infrastructure changes that alter Azure Machine Learning workspace private-endpoint settings before those changes reach production. What is the best control?
- A clerk's office forbids long-lived personal access tokens for Azure Machine Learning and GitHub integration and prefers a short-lived federated identity pattern. What should they configure?
- A school-district MLOps runbook must apply Bicep to create a workspace and then register default environments in that workspace during the same automation. Which approach best completes both steps?
- A water utility security review finds private endpoints on an Azure Machine Learning workspace, yet the workspace still allows public network access. What must the team enforce to meet the review?
- A multi-agency grant project needs experiment pipeline definitions kept separate from production definitions, with required reviews before production changes. Which source-control practice should they use?
- After a Bicep deploy, a city SRE wants CI to confirm the Azure Machine Learning workspace managed identity can reach the registered datastore before training jobs are allowed. What should CI include?
- A county wants ephemeral Azure Machine Learning experiment workspaces removed automatically after short-lived grant pilots to control cost. Which approach best supports that lifecycle?
Implement machine learning model lifecycle and operations · 56 questions
- A permits office wants each model experiment run to record parameters and metrics inside Azure Machine Learning for later comparison. Which capability should they configure?
- A public-health team must choose among MLflow-tracked runs for a disease-outbreak early-warning model by comparing AUC and recall. What should they do?
- A transit authority wants to explore many candidate algorithms quickly for next-day ridership prediction before committing to a hand-tuned pipeline. Which Azure Machine Learning approach fits best?
- A housing authority must keep AutoML within interpretable model classes so fair-housing review boards can explain decisions. How should AutoML be configured?
- A city data scientist needs to prototype feature ideas for an illegal-dumping classifier before the team builds a production training pipeline. Which approach is most appropriate for that early stage?
- A water utility wants systematic search over hyperparameters for a leak-detection model instead of overnight hand-tuning. What should the MLOps engineer configure?
- A fire-district team needs to run a smoke-plume estimation training script on Azure Machine Learning with arguments captured for audit. Which approach should they use?
- A county assessor must train a large imagery model that no longer fits practical single-node GPU limits. What should the team use in Azure Machine Learning?
- A 311 operations group needs data-prep, train, and evaluate steps to run as a repeatable nightly refresh rather than ad hoc notebook clicks. What should they implement?
- A parks department wants to decide which tree-canopy classifier generalizes better on holdout wards by reviewing two training jobs together. What should they do?
- A clerk's office wants scikit-learn training scripts to record metrics automatically in MLflow so interns are not responsible for manual screenshots. What should they enable?
- A municipal fraud unit wants AutoML to handle messy grant-application tables while preserving careful train and validation splits. Which practice should guide the AutoML run?
- A school-district energy team documented thermostat setpoint model assumptions in an Azure ML notebook and now needs governed training for production. What should they do next?
- A transit MLOps engineer needs to search learning rate and tree depth for a delay-classification model without hand-picking every trial. Which Azure ML approach should they configure?
- A public-works crew wants its forecasting model trained by a reusable unit that wraps the training script inside municipal pipelines. What should they use?
- A regional emergency ops center finds its wildfire-risk CNN exceeds single-GPU memory during training. What should the MLOps team do?
- A city revenue office retrains a delinquency-risk model every quarter and wants the same orchestration each cycle. Which pattern should they implement?
- A library consortium must confirm which data asset version produced the winning catalog-tagging model before approving it. What should reviewers examine?
- A courts analytics pilot needs fairness measures recorded with accuracy for pretrial risk experiments. How should the team capture those measures in Azure ML?
- A sanitation department wants many hyperparameter trials for a route-optimization model but must control Azure spend. Which sweep practice should they apply?
- A permits MLOps lead selected a winning experiment run and needs it available for controlled production promotion. What should they do in Azure Machine Learning?
- A public-health team wants scoring to pull the same features used at training time for an outbreak early-warning model. What should they package with the model artifact?
- A transit authority is ready to promote a ridership model into the registry but must check responsible AI concerns first. What should they do before promotion?
- A housing authority retired an eligibility-scoring model and must prevent accidental redeployment. Which lifecycle action should they take?
- A city fraud team is registering grant-screening model v3 while grant season still may require a fast return to v2. How should versions be handled?
- A water utility is registering a quality-prediction MLflow model and needs reliable scoring contracts for downstream systems. What should registration include?
- A county assessor is promoting a valuation-assist model and governance requires evidence of fairness and error analysis. What should accompany the promotion ticket?
- A 311 team must ensure production enrichment uses the same feature store retrieval path the registered model expects. What should the registered model reference?
- A parks department ends a failed tree-canopy experiment but auditors may still need the model history. What is the appropriate registry action?
- A clerk's office requires clear promotion states before any endpoint deploy of registered models. Which practice should their MLOps runbook enforce?
- A fire district's responsible AI evaluation shows disparate error rates across neighborhoods for a candidate dispatch-aid model. What should the promotion decision be?
- A school-district MLOps runbook needs batch scoring to stay consistent with training features for an enrollment model. How should artifacts be packaged?
- A city needs instant permit-risk scores while clerks enter applications at the counter. Which deployment option should MLOps choose?
- A transit authority must score next-day ridership for all routes overnight without interactive latency requirements. Which endpoint pattern fits?
- A public-health team is about to cut over clinic traffic to a new online endpoint. What should they do first?
- A housing authority online endpoint returns HTTP 429 responses during peak voucher-application hours. What is the most likely operational issue to address?
- A water utility wants to release a new SCADA-derived scoring model version with limited blast radius. Which release strategy should they use?
- A 311 platform swaps in a new triage model and latency SLOs start failing. What should the MLOps team do first to protect residents?
- A courts analytics unit needs a document-risk classifier in production without babysitting unmanaged virtual machines. Which deployment approach fits best?
- A parks mobile app fails scoring because the endpoint environment is missing a geospatial dependency. How should ops isolate the fault?
- A county revenue office wants to cut over a new parcel-scoring batch endpoint safely. What progressive practice should they use?
- A fire-district dispatch aid model must stay responsive during incident spikes. Which online endpoint capability should they configure?
- A library catalog-tagging service must catch broken request payloads before go-live. What should endpoint tests verify?
- A city MLOps policy covers progressive rollout of resident-facing scores. What operational readiness item is required before traffic shifts?
- After a permits intake form redesign, feature distributions shift and the live model’s inputs no longer resemble training. What should monitoring detect?
- A transit ridership model is live and delayed-trip labels arrive each week. What should the team track in production?
- A public-health outbreak model must warn leadership when quality falls. Which monitoring control fits?
- A housing authority agrees that statistical drift beyond a set distance should refresh the eligibility model. What should they configure?
- A water-quality scoring service sees drift signals and must decide whether to retrain. What analysis should happen first?
- A 311 classifier dashboard shows rising latency and falling F1 after ticket categories change. What monitoring insight does this illustrate?
- A courts risk model’s drift panels must respect sealed-case constraints. How should monitors be configured?
- A parks visitation model breaches the municipal endpoint error-rate SLO. Which alert type should fire for on-call?
- A school-district enrollment forecaster only learns true fall headcount months later. How should production quality be monitored?
- A fire-district aid model receives drift alerts but unstable automatic updates would be dangerous. What retrain policy fits?
- A city open-data MLOps team needs drift detection that auditors can defend. What should they document?
- A clerk’s office must approve model promotions using production evidence. Which practice supports that governance?
Design and implement a GenAIOps infrastructure · 46 questions
- A city GenAI team is building a permitting chatbot and already has a classical Azure Machine Learning workspace. What should they create for GenAIOps?
- A county Foundry project must read municipal knowledge from Blob storage without embedding secrets in configs. What should they configure?
- A public-health GenAI app lets clinicians invoke agents but must stop them from changing foundation-model deployments. Which control applies?
- A transit authority handles rider-PII prompts through Foundry and must keep that traffic off the public internet. What network posture should they implement?
- A housing authority must provision Foundry infrastructure consistent with the city’s landing-zone standards. Which approach fits?
- A 311 program needs the same Foundry project layout in staging and production subscriptions. What scripting approach helps?
- A courts GenAI pilot must keep experimental prompts away from resident-facing production agents. What Foundry approach helps?
- A water utility needs operators and prompt engineers to have different Foundry permissions. What should platform admins do?
- A library consortium must meet the city security baseline that forbids public internet exposure on Foundry-dependent resources. What should they enforce?
- A parks chatbot platform proposes Bicep changes that add Foundry connections to municipal SharePoint knowledge bases. What review practice fits?
- A school-district GenAI lab wants classroom assistants under one governed Foundry resource without mixing every pilot into a single flat project. What structure helps?
- A fire-district assistant must call Azure AI Search grounding indexes without embedding secrets in the app. Which identity pattern fits?
- A city SRE must confirm Foundry-dependent model endpoints are reachable only through the private network before go-live. What should they validate?
- A clerk’s office wants Foundry provisioned by Azure CLI, then access controls applied as a separate governed step. How should the pipeline be staged?
- A municipal AI CoE must allocate Foundry cost across departments and keep inventories searchable. What resource hygiene helps?
- A county security team wants GenAI builders productive without granting blanket Owner on Foundry. What access approach fits?
- A permits chatbot sees spiky daytime traffic and quiet nights. Which foundation-model hosting choice fits variable load?
- A transit multimodal assistant must run a vision-capable foundation model on station camera stills with controlled compute. What deployment option fits better than pure bursty serverless chat?
- A public-health team needs a clinical FAQ assistant and must avoid oversized general chat models that overspend. What selection approach fits?
- A housing authority runs prompt evaluations that must target the exact foundation-model revision used in staging. What practice enables that?
- A 311 program expects a storm-driven surge and must keep answer latency inside SLA. Which capacity control helps?
- A courts summarization service has a new foundation-model version ready. What production deployment strategy fits?
- A water-utility ops copilot must generate SCADA query helpers. Which foundation-model selection fits better than a generic creative chat model?
- A library chat pilot is early and traffic is uncertain. How should they approach provisioned throughput?
- A parks recreation assistant must meet city policy that requires content-filter-compatible foundation-model settings. What should deployment include?
- A school-district tutoring bot must only serve a foundation-model version that already passed the district evaluation suite. What should operators do?
- A fire-district after-action report assistant finds an oversized foundation model that exceeds the city’s per-token budget. What should the team do?
- A city open-data narrator expects a council-meeting livestream Q&A spike. How should PTU capacity be handled?
- A clerk’s generative forms helper must redraft large overnight batches, but serverless quotas cannot meet the window. What hosting choice fits?
- A county GenAIOps board requires a canary week when updating foundation models. What deployment pattern satisfies that policy?
- A sanitation route-advice agent serves field workers on mobile devices where response delay hurts job flow. What model-selection priority fits?
- A municipal contact center wants reserved capacity for the production Foundry deployment while keeping lab experiments cheaper. What should GenAIOps do?
- A permits chatbot must stop inventing ordinance language and instead cite city code. What prompt-design practice helps most?
- A 311 assistant team wants to try concise versus detailed system prompts and compare how residents react. What should they do?
- A public-health FAQ bot must not push prompt edits straight to production Foundry. What control fits GenAIOps?
- A transit disruption bot’s diffs are messy because system instructions and few-shot examples live in one blob. What structuring practice helps?
- A housing eligibility helper must choose between two prompt variants before go-live. What comparison approach is sound?
- A courts clerk assistant may only run prompts that auditors can prove were approved. What Git practice supports that?
- A water-quality advisor must not invent medical advice when residents ask about illness after a boil-water notice. What prompt redesign helps?
- A parks event bot serves residents in several languages and must check whether prompt variants stay equally helpful. What should GenAIOps do?
- A school-district GenAIOps pipeline must stop production when Foundry prompts no longer match source control. What enforcement fits?
- A library research assistant must not paste entire copyrighted books into chat. What prompt-design addition helps?
- A fire-public-info bot has experimental prompt branches that must never reach the live Foundry deployment slot. What practice keeps them apart?
- A city grants office is comparing prompt variants that change how retrieval instructions are written. Which metric comparison is most appropriate?
- A clerk’s office needs auditors to understand who wrote each prompt and why it exists. What Git hygiene helps?
- A municipal AI board finds a clever prompt that strips required disclaimer language from resident legal questions. What should happen?
Implement generative AI quality assurance and observability · 30 questions
- A permits chatbot team needs a fair way to judge answer quality before each release. What evaluation asset should they build?
- A 311 evaluation harness must check whether answers stick to the city’s published service catalog. Which quality metric fits best?
- A public-health assistant answering vaccination-hours questions needs more than a single vague score. What evaluation approach fits?
- A housing chatbot receives free-text dialogs that might include harmful or self-harm content. What evaluation should GenAIOps configure?
- A transit agent pipeline must re-check quality whenever prompts or models change. What workflow should GenAIOps set up?
- A courts summarizer must be scored on whether required summary fields appear from docket excerpts. What test-dataset practice helps?
- A water-advisor bot’s answers become jumbled after a prompt edit even though some facts remain. What quality gate should block promotion?
- A library youth-services bot must align safety checks with age-appropriate content policies. What should GenAIOps configure?
- A parks recreation assistant must quote fees correctly from the fee schedule, not only sound fluent. What evaluation addition helps?
- Teachers find edge-case word problems that the school-district tutoring bot mishandles, and those cases are missing from eval. What should GenAIOps do?
- A fire-public-info agent answers evacuation-zone questions grounded in GIS bulletins. Which quality metric best checks that answers address the resident’s location-critical ask?
- A city contact center must not ship a municipal policy bot when safety evaluations show elevated jailbreak success. What release practice fits?
- A clerk’s generative forms helper must fail the release pipeline when Foundry evaluation metrics regress. How should GenAIOps wire that gate?
- A grants office needs Spanish and English prompts to share equivalent expected intents in the evaluation set. What dataset practice supports fair bilingual scoring?
- A sanitation chatbot’s fluency scores look fine while groundedness collapses on pickup-schedule facts. What should ops prioritize before promotion?
- A municipal AI review board requires evaluation evidence in every GenAI go-live packet. Which package meets that governance need?
- A 311 Foundry deployment just launched and product owners need live quality signals after go-live. Which capability should they enable?
- A transit assistant must meet rider-app SLOs for how quickly and how many requests it handles. Which observability focus fits?
- A public-health chatbot must stay inside clinic conversation budget limits. What should ops track?
- A housing agent fails eligibility checks when tool calls break, and support needs to debug production. What observability capability helps most?
- After a courts corpus refresh, continuous monitoring shows a spike in low-groundedness responses. What is the appropriate ops interpretation?
- A water-ops copilot dashboard shows p95 latency rising during morning SCADA report generation. Why does that percentile matter?
- A library chatbot cost report shows higher token burn after a verbose system prompt shipped last week. How should cost metrics guide the fix?
- A parks event bot needs support tickets linked to GenAI traces for faster triage. What logging practice enables that?
- A school-district tutoring service expects exam-week demand spikes and wants to avoid throttling. What should observability emphasize beforehand?
- A fire-district assistant’s traces show repeated retrieval timeouts against the policy index. What should ops conclude?
- A city AI CoE wants department product owners engaged with Foundry continuous monitoring. Which operational practice fits?
- A clerk’s office must know when Foundry token spend exceeds the monthly appropriation. What cost-control practice fits?
- A sanitation route advisor samples production traces and finds possible PII in logs. What privacy-aware observability step fits?
- A grants chatbot SLO pack must reflect both resident wait time and spend efficiency. Which combination belongs in the pack?
Optimize generative AI systems and model performance · 30 questions
- A permits RAG bot returns irrelevant code sections for resident questions. What optimization should engineers try first for retrieval quality?
- A 311 knowledge base of long PDFs retrieves noisy chunks that miss the right procedure. Which chunking change often improves civic RAG?
- A public-health RAG system struggles with clinical-guideline wording in the city’s corpus. What embedding approach fits domain accuracy?
- A housing policy assistant must find exact program names while still supporting semantic questions. Which retrieval approach fits?
- A transit disruption bot team wants evidence that a new retrieval strategy beats the current one. What evaluation approach fits?
- A courts RAG tool keeps citing weakly related opinions that confuse clerks. Engineers want fewer weakly grounded answers without rewriting the whole index. What should they tune first?
- A water-quality FAQ struggles with treatment-plant jargon in SOP PDFs even though chunking looks fine. What embedding change best improves domain retrieval?
- A library catalog assistant finds topical books well but often misses exact ISBN lookups residents paste into chat. What retrieval approach should GenAIOps add?
- A parks permits RAG team changes how ordinance PDFs are chunked and needs proof retrieval quality improved. What should they use?
- A school-district policy bot fails on multi-part parent questions that a single naive top-k query cannot cover. What retrieval strategy change fits best?
- A fire-code assistant often retrieves the wrong code sections despite decent chunking. Annotated pairs of queries and correct sections are available. What should the team do to improve lookup accuracy?
- A city grants portal RAG misses exact NOFO identifiers that applicants type precisely, even when semantic search finds related topics. What should GenAIOps implement?
- A clerk’s ordinance bot team wants a fair comparison of two rerankers before changing production RAG. What evaluation approach should they use?
- A sanitation SOP assistant’s answers often miss steps that span adjacent sections because chunks are too small. What retrieval tuning should the team try?
- A municipal open-data narrator must reliably retrieve ward names and local toponyms in resident questions. How should the team choose among embedding models?
- A 311 contact-center GenAIOps team claims a hybrid index beat pure vector search for resident FAQs. What should leadership require before accepting the change as proven?
- A permits office wants a city-tone drafting model customized on limited GPU budget without full foundation-model retraining. Which fine-tuning approach fits best?
- A 311 team lacks enough labeled resident chats to fine-tune an intent-routing helper. What data strategy should they use?
- A public-health team fine-tunes a vaccination-FAQ model and must catch quality regressions across checkpoints. What should they do after each training checkpoint?
- A housing authority finished sandbox evaluation of a fine-tuned eligibility explainer and is ready for residents. What lifecycle step should GenAIOps take next?
- A transit authority needs to customize a foundation model for agency acronyms and stop names without full-weight retraining. Which method best fits?
- A courts clerk team needs fine-tuning examples that resemble dockets but must not train on sealed real cases. What should they create?
- A water-utility fine-tuned assistant shows rising toxic-content flags on safety evaluations after a new checkpoint. What should ops do first?
- A library summarizer fine-tune is ready, but production still serves the prior model. What promotion sequence is correct?
- A parks department wants a copywriting model that follows approved park-board style guides rather than generic marketing tone. Which fine-tuning design fits?
- A school-district tutoring fine-tune uses synthetic math word problems that change over time. What should GenAIOps do with those datasets?
- A fire-district after-action assistant was fine-tuned and must prove it improved over the base foundation model before cutover. What comparison should ops run?
- A city GenAIOps runbook is cutting over to a new fine-tuned checkpoint that cleared eval and canary. What should happen to the old fine-tuned deployment?
- A grants office is choosing between continued pretraining and chat-style fine-tuning for NOFO-heavy drafting. How should they decide?
- A clerk’s office runs a fine-tuned forms helper and must decide whether customization still pays off. What monitoring loop should GenAIOps use?
These questions are original practice material and are NOT actual exam questions or brain-dump content. All vendor marks are trademarks of their respective owners. This site is not affiliated with, endorsed by, or sponsored by Microsoft Corporation.