Skip to content

Topic Drill 1.1.6: Development and Hosting of AI Models ​

About this drill

LO AI-1.1.6 (K2): Compare the options for the development and hosting of AI models.
Lesson: Ch 1, section 1.1.6

This drill covers every point in section 1.1.6:

  • Third-party (pretrained models, AIaaS: faster time-to-market) vs. private (better tailored, specialized skills)
  • Local development by model size; public vs. private cloud development; hybrid development
  • Local hosting (PCs, smartphones, dedicated servers); public vs. private cloud hosting; hybrid hosting
  • Decision factors, separate decisions for development and hosting, multi-tiered strategies

Topic Drill: 1.1.6

Q1. A start-up with no ML specialists must launch a customer-support chatbot within six weeks. Which option does the syllabus suggest fits BEST, and why?

AA private AI-based system, because it can be better tailored to their requirements
BLocal development on on-premises server clusters, because it gives direct control
CA third-party pretrained model or AI as a Service (AIaaS), because it enables faster deployment and quicker time-to-market
DTraining a new model from scratch on a public cloud, because pay-as-you-go pricing is cheap

Q2. A defence contractor needs a model tailored to very specific requirements and employs a team of ML experts. Which option does the syllabus associate with this situation?

AA private AI-based system (on-premises or customized cloud), which can be better tailored but requires specialized skills
BAn off-the-shelf AIaaS offering, because it is quicker to market
CLocal hosting on smartphones, because it eliminates licensing costs
DA public cloud with pay-as-you-go pricing, because it limits initial investment

Q3. According to the syllabus, what is needed for LOCAL development of large-scale models?

AA personal computer with a fast CPU
BA single dedicated GPU
CA smartphone with an AI chip
DHigh-performance on-premises server clusters, with their associated energy, cooling and hardware costs

Q4. Which set of advantages does the syllabus give for developing on a PUBLIC cloud?

AEnhanced security and privacy, with a greater upfront investment
BPre-configured environments and pay-as-you-go pricing, which limit initial hardware investment and scale easily
CDirect control and privacy of local coding and training
DNo need to consider regulatory requirements

Q5. A health-insurance company chooses a private cloud for model development. What trade-off does the syllabus describe for this choice?

ALower cost but less control
BNo maintenance, but limited scalability
CEnhanced security and privacy, at the price of a greater upfront infrastructure investment
DFaster time-to-market, but no possibility of customization

Q6. A note-taking app runs a small summarization model directly on users' phones. Which benefits and limitation does the syllabus give for this kind of local hosting?

APrivacy and no cloud licensing costs, but limited hardware capabilities
BUnlimited scalability, but high licensing costs
CEnhanced control, but a significant upfront investment in dedicated servers
DNo maintenance concerns, which makes it ideal for fluctuating workloads

Q7. An online ticket shop sees huge demand peaks when concerts go on sale and very little traffic otherwise. Which hosting option does the syllabus describe as ideal for such fluctuating workloads?

ALocal hosting on employees' laptops
BDedicated on-premises servers
CA single PC in the shop's office
DA public cloud, offering scalable access to robust infrastructure without maintenance concerns

Q8. Which TWO factors does the syllabus list as influencing the choice of development and hosting solutions?

Select 2 options.

AThe brand of the developers' IDE
BModel size
CThe number of testers in the team
DThe popularity of the programming language
ERegulatory requirements

Q9. Which statement about development and hosting decisions is CORRECT according to the syllabus?

ADevelopment and hosting must always use the same environment
BThe optimal development and hosting solutions are typically decided upon separately
CHosting decisions are made by the model during training
DHybrid approaches are only allowed for hosting, not for development

0 / 9 answered

← 1.1.5 Hardware for ML Systems · Next drill: 1.1.7 ML Development Frameworks →

Unofficial study guide for the ISTQB® Certified Tester AI Testing (CT-AI) exam. Not affiliated with International Software Testing Qualifications Board (ISTQB®).