Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Cloud Digital Leader practice bank has an additional 3,070 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
A specialized electronics manufacturer wants to gain a competitive advantage by automating real-time defect detection on its assembly line. The company has a following set of operational circumstances and technical requirements:
Which Vertex AI approach should the company use to address its unique business needs and differentiate its quality control process?
Model distillation (also known as knowledge distillation) is a machine learning compression and optimization technique that uses a teacher-student framework. In this process, a smaller, more compact "student" model is trained to mimic the predictions and behavior of a much larger, complex "teacher" model.
This option would be correct if the organization already had a massive, highly accurate foundation model performing well on a task, and their primary goal was to create a smaller, faster, and more cost-effective version of that model to run on devices with limited compute resources.
Model distillation is a powerful technique within Vertex AI for optimizing deployment costs and serving latency, especially for large language models, but it is an optimization strategy rather than a substitute for training a custom model on unique proprietary data.
Custom model training in Vertex AI is the process of building, training, and deploying machine learning models tailored to an organization's specific business problems using its own unique datasets. Unlike generic off-the-shelf solutions, custom training gives data scientists and machine learning engineers complete control over the underlying model architecture, training code, and optimization parameters.
While pre-trained APIs and prompt-engineered foundation models offer rapid deployment for general use cases, they lack the specialized precision and architectural control needed for proprietary industrial inspection. Building a custom model on Vertex AI transforms the organization's unique historical data into a strategic asset that directly improves product quality and operational efficiency.
Prompt engineering (or prompt design) is the practice of crafting natural language instructions, context, and examples within a prompt to guide a generative AI foundation model toward producing a desired output. Because foundation models are pre-trained on multi-terabyte datasets, they exhibit emergent abilities to perform varied tasks without modifying the model's internal weights.
This approach would be appropriate if the company lacked a labeled training dataset, prioritized the fastest possible go-to-market (GTM) timeline over model training control, or needed to generate written inspection summaries and maintenance reports from general inputs.
Even when organizations use generative AI for specialized tasks, prompt engineering alone is often not enough; they typically must progress to supervised tuning using labeled data examples. For strict control over algorithms and hyperparameters in defect classification, custom training remains necessary.
Pre-trained traditional AI models—such as Vision AI, Document AI, Natural Language API, and Video Intelligence API—are ready-to-use services trained by Google on massive, general-purpose datasets. They enable developers to extract insights from images, text, and video through simple API calls without needing machine learning expertise or training data.
This option would be the right choice if the organization needed to perform standard visual tasks—such as Optical Character Recognition (OCR) to read serial numbers on shipping labels or basic object detection of common items—where a pre-trained model already meets the use case requirements and rapid deployment is the primary goal.
In a standard machine learning decision workflow, organizations should always evaluate whether a pre-trained model meets their requirements first. However, once it is clear that generic models cannot handle the specialized domain task and sufficient proprietary data is available, moving to custom model training in Vertex AI is the appropriate strategic step.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated