引導式測驗 · 中等 等級

人工智慧基礎設施工程師職業測驗

測試基礎設施層、雇主範例、工程權衡以及從標題推斷範圍或規模的限制。

問題 1 的 8

Which responsibility bundle is specifically listed in Microsoft’s Compute Infra/HPC posting?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. Which responsibility bundle is specifically listed in Microsoft’s Compute Infra/HPC posting?
    • Develop model architectures and training recipes without infrastructure ownership
    • Own product requirements and customer roadmaps for AI features
    • Operate a user-facing API without managing compute or model lifecycle systems
    • Build compute control planes, provision clusters, manage fleet health, and automate operations
  2. Which model-lifecycle span does Microsoft AI Infrastructure and Model Foundry describe supporting?
    • Experimentation and training through evaluation, packaging, deployment, inference, and observability
    • Model architecture research only, with no experimentation, deployment, or inference platform
    • Build user-facing application features without owning model lifecycle systems
    • Procure accelerator hardware while another group owns all software and operations
  3. For a production model-serving system, which tradeoff bundle does the Microsoft posting name?
    • Model novelty and benchmark scores, without reliability or operating cost
    • Developer adoption and product engagement, without system performance measures
    • Latency, throughput, reliability, and infrastructure cost
    • Training accuracy and convergence, without serving or capacity measures
  4. Which detail do the cited postings not establish as a universal requirement for AI infrastructure roles?
    • That infrastructure engineers may work across distributed systems and hardware
    • An exact accelerator count or cluster size that every employer must use
    • That one cited role works on control-plane and cluster systems
    • That one cited role works on model serving and accelerator orchestration
  5. Which example changes an infrastructure property without changing the model’s research method?
    • Choose new labels and redesign the training data collection study
    • Compare scientific explanations for a model behavior
    • Profile the serving path and improve inference throughput or recovery behavior
    • Propose a new learning objective and evaluate a research hypothesis
  6. Which capability set does the guide identify as useful across infrastructure specialties?
    • Distributed-systems reasoning, debugging, reliability, and performance analysis
    • Model architecture, experimental design, and research publication
    • Roadmap prioritization, customer discovery, and launch communication
    • Board design, semiconductor physics, and chip verification
  7. Which scope question would best clarify an infrastructure team’s day-to-day work?
    • Which cloud or accelerator vendor does the team support?
    • Which system layer the role owns, and how much work is platform building versus operations
    • Which programming language is used in the team’s codebase?
    • Which model family does the product use?
  8. What should readers avoid concluding from the two Microsoft job examples?
    • One posting focuses on compute control planes and fleet health
    • A separate posting covers model-serving infrastructure
    • Current job descriptions can reveal team-specific expectations
    • Every AI infrastructure role uses the same hardware scale and performs the same duties