How should I use these Architecting Low-Code ML Solutions practice questions?

Read each scenario carefully and choose your answer before revealing the explanation. Then check why your choice was right or wrong. Repeat until the reasoning feels automatic.

Can I practise just Architecting Low-Code ML Solutions questions in a focused session?

Yes — use the session launcher on this page to start a 10-, 20-, 30- or 50-question session drawn entirely from the Architecting Low-Code ML Solutions domain.

PMLE · topic practice

Architecting Low-Code ML Solutions practice questions

This domain covers building ML with minimal code using Google Cloud managed services. It tests BigQuery ML model selection (LINEAR_REG, LOGISTIC_REG, BOOSTED_TREE), pretrained AI APIs like Natural Language and Document AI, and when to add human review. Expect scenario questions choosing the right service for structured prediction, document extraction, or sentiment analysis.

Courseiva uses original exam-style practice questions designed for learning and revision. The goal is to understand the concepts, recognise exam patterns, and improve through explanations — not memorise copied exam dumps.

Editorial oversight:Johnson Ajibi· MSc IT Security, IEEE Senior Member

20 questionsDomain: Architecting Low-Code ML Solutions

Practice 10 questions Browse domain →

What the exam tests

What to know about Architecting Low-Code ML Solutions

Be able to pick the right low-code service: BigQuery ML model type by prediction task, pretrained API by data modality, and human review for high-risk cases. The most important thing is matching the service to the required output and risk level.

Selecting BigQuery ML model types: LINEAR_REG for regression, LOGISTIC_REG for binary classification, BOOSTED_TREE for automatic tuning

Using Document AI to extract structured text fields and tables from scanned PDFs

Applying Natural Language API for sentiment analysis with minimal customization

Choosing human-in-the-loop review for high-risk automated decisions in financial workflows

Watch out for

Common Architecting Low-Code ML Solutions exam traps

▸Assuming BigQuery ML requires exporting data or writing Python; it trains models directly with SQL CREATE MODEL statements
▸Confusing pretrained API use cases: Natural Language for sentiment versus Document AI for PDF form and table extraction
▸Forgetting that high-risk loan or financial decisions need a human review step, not fully automated API output

Practice set

Architecting Low-Code ML Solutions questions

20 questions · select your answer, then reveal the explanation

Question 1hardmultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A company wants to use BigQuery ML to train a DNN_CLASSIFIER model on a dataset with 100 million rows. They are concerned about training time and cost. Which approach can help optimize training performance while staying within BigQuery ML?

Trap 1: Use OPTIONS('MAX_ITERATIONS' = 10) to limit training iterations

While this can reduce time, it may result in underfitting; not an optimization recommendation.

Trap 2: Use Vertex AI AutoML Tables instead of BigQuery ML

This introduces additional complexity and data export; the question asks to stay within BigQuery ML.

Trap 3: Train on a 10% random sample of the data to reduce cost

Sampling reduces representativeness and may not be acceptable for production models.

Study all Architecting Low-Code ML Solutions common traps →

A
Use OPTIONS('MAX_ITERATIONS' = 10) to limit training iterations
Why it fails: While this can reduce time, it may result in underfitting; not an optimization recommendation.
B
Use Vertex AI AutoML Tables instead of BigQuery ML
Why it fails: This introduces additional complexity and data export; the question asks to stay within BigQuery ML.
C
Train on a 10% random sample of the data to reduce cost
Why it fails: Sampling reduces representativeness and may not be acceptable for production models.
D
BigQuery ML automatically optimizes training; no additional configuration needed
BigQuery ML handles optimization internally, adjusting training parameters for efficiency.

Full breakdown with real-world context →

Question 2mediummultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A company needs to analyze customer feedback from app reviews to identify common themes and sentiment. They have millions of reviews in multiple languages. Which combination of pre-built APIs should they use?

Trap 1: Speech-to-Text then Natural Language API

App reviews are text, not audio.

Trap 2: Natural Language API only

Natural Language API supports multiple languages for sentiment analysis, but for best accuracy, translation may be needed if reviews are in many languages.

Trap 3: Vision API then Natural Language API

Vision API is for images; app reviews are text.

Study all Architecting Low-Code ML Solutions common traps →

A
Speech-to-Text then Natural Language API
Why it fails: App reviews are text, not audio.
B
Natural Language API only
Why it fails: Natural Language API supports multiple languages for sentiment analysis, but for best accuracy, translation may be needed if reviews are in many languages.
C
Translation API then Natural Language API
Translate reviews to a common language, then analyze sentiment and entities.
D
Vision API then Natural Language API
Why it fails: Vision API is for images; app reviews are text.

Full breakdown with real-world context →

Question 3mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A retail company uses Recommendations AI to power personalized product recommendations on their website. They notice that the 'frequently-bought-together' model is not capturing complementary items that are often purchased in the same session but not necessarily in the same transaction. Which TWO actions should they take to improve the model?

Trap 1: Decrease the event retention period to focus on recent purchases

This would reduce data and likely harm model quality.

Trap 2: Use AutoML Tables to build a custom recommendation model

AutoML Tables is not designed for recommendation systems; it would require significant custom effort.

Trap 3: Set the recommendation type to only 'frequently-bought-together'

This limits to transaction-level co-purchase, not session-level.

Study all Architecting Low-Code ML Solutions common traps →

A
Decrease the event retention period to focus on recent purchases
Why it fails: This would reduce data and likely harm model quality.
B
Enable the 'others-you-may-like' recommendation type in addition to 'frequently-bought-together'
'Others-you-may-like' uses co-viewed and co-purchased signals, capturing session-level patterns.
C
Use AutoML Tables to build a custom recommendation model
Why it fails: AutoML Tables is not designed for recommendation systems; it would require significant custom effort.
D
Set the recommendation type to only 'frequently-bought-together'
Why it fails: This limits to transaction-level co-purchase, not session-level.
E
Ingest session-level event data (e.g., product views in the same session) into Recommendations AI
Session-level events capture co-viewed items across sessions, improving recommendations.

Full breakdown with real-world context →

Question 4mediummultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A retail company wants to build a recommendation system for their e-commerce website. They have user purchase history and product metadata. Which Google Cloud service is most suitable for building a 'frequently bought together' recommendation model with minimal custom ML development?

Question 5mediummultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A retailer wants to implement a recommendation engine that suggests products based on a user's current cart. They have limited ML expertise and want a quick deployment. Which Recommendations AI model type should they use?

Question 6easymultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A company needs to detect objects in real-time from a live video feed. They want to use a pre-trained model with minimal setup. Which Google Cloud service should they use?

Question 7mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A data scientist wants to use BigQuery ML for time-series forecasting. They need to evaluate model accuracy and compare different models. Which TWO BigQuery ML functions should they use?

Trap 1: ML.EXPLAIN_PREDICT

ML.EXPLAIN_PREDICT provides local explanations for predictions, not evaluation or comparison.

Trap 2: ML.FEATURE_IMPORTANCE

ML.FEATURE_IMPORTANCE is used for tree-based models to show feature importance, not for evaluation.

Trap 3: ML.TRAIN

ML.TRAIN is not a valid BigQuery ML function; model training is performed via CREATE MODEL statement.

Study all Architecting Low-Code ML Solutions common traps →

A
ML.EXPLAIN_PREDICT
Why it fails: ML.EXPLAIN_PREDICT provides local explanations for predictions, not evaluation or comparison.
B
ML.PREDICT
ML.PREDICT generates forecasts, which are necessary for comparing model outputs.
C
ML.FEATURE_IMPORTANCE
Why it fails: ML.FEATURE_IMPORTANCE is used for tree-based models to show feature importance, not for evaluation.
D
ML.EVALUATE
ML.EVALUATE directly computes evaluation metrics like MAE, MSE, etc., enabling model comparison.
E
ML.TRAIN
Why it fails: ML.TRAIN is not a valid BigQuery ML function; model training is performed via CREATE MODEL statement.

Full breakdown with real-world context →

Question 8mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A company needs to classify images of products into categories (e.g., electronics, clothing, food). They have labeled images and want to use a low-code solution on Google Cloud. Which service is suitable for this task?

Trap 1: AutoML Tables

AutoML Tables is designed for tabular data, not image classification.

Trap 2: Cloud Vision API using product search

Cloud Vision API product search is for finding similar products, not custom classification into categories.

Trap 3: Vertex AI Workbench with a custom TensorFlow model

Vertex AI Workbench with a custom TensorFlow model requires writing custom code, so it is not low-code.

Study all Architecting Low-Code ML Solutions common traps →

A
AutoML Tables
Why it fails: AutoML Tables is designed for tabular data, not image classification.
B
AutoML Vision
AutoML Vision provides a low-code, UI-driven workflow for training custom image classification models on labelled images, satisfying the requirement to categorise products without writing model code. It handles electronics, clothing and food categories directly from the labelled dataset.
C
Cloud Vision API using product search
Why it fails: Cloud Vision API product search is for finding similar products, not custom classification into categories.
D
Vertex AI Workbench with a custom TensorFlow model
Why it fails: Vertex AI Workbench with a custom TensorFlow model requires writing custom code, so it is not low-code.
E
Document AI custom extractor
Why it fails: Document AI is specialized for extracting data from documents, not general image classification.

Full breakdown with real-world context →

Question 9mediummultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A company wants to transcribe customer service calls in real-time. The audio is telephony quality (8 kHz). Which Speech-to-Text model should they specify?

Question 10hardmultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A company uses BigQuery ML with a remote model calling Vertex AI's pre-trained image classification model. They need to classify images stored in Cloud Storage buckets. What is the correct approach?

Question 11mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A data scientist wants to use AutoML Tables to build a binary classification model for loan default prediction. The dataset has 200 features and 1 million rows, with highly imbalanced classes. Which TWO options should they consider? (Choose 2)

Question 12mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A company needs to detect objects in live video streams from security cameras. They require low-latency predictions and want to minimise operational overhead. Which TWO services should they use? (Choose 2)

Question 13hardmulti select

Read the full Architecting Low-Code ML Solutions explanation →

A company wants to use Document AI to process a large volume of invoices. They need to extract line items and also have a human review the extracted data for accuracy. Which THREE features should they use? (Choose 3)

Question 14hardmulti select

Read the full Architecting Low-Code ML Solutions explanation →

A data engineer is using BigQuery ML with a BOOSTED_TREE_CLASSIFIER model. After training, they want to evaluate the model and understand which features contribute most to predictions. Which THREE BigQuery ML functions should they use?

Question 15mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A company needs to build a custom model to classify images of products into categories. They have a large labeled dataset. They want to use AutoML but are unsure which options support image classification. Which TWO AutoML products support image classification?

Trap 1: AutoML Natural Language

AutoML Natural Language is for text, not images.

Trap 2: AutoML Translation

AutoML Translation is for language translation, not images.

Trap 3: AutoML Tables

AutoML Tables is for structured tabular data, not images.

Study all Architecting Low-Code ML Solutions common traps →

A
AutoML Natural Language
Why it fails: AutoML Natural Language is for text, not images.
B
AutoML Video
AutoML Video Intelligence can classify images in video frames, making it valid for image classification.
C
AutoML Vision
AutoML Vision is Google Cloud's managed image classification and object detection service, so it directly satisfies the requirement to train a custom model on the large labelled image dataset. AutoML Video handles video, not still images, making Vision the correct product for this scenario.
D
AutoML Translation
Why it fails: AutoML Translation is for language translation, not images.
E
AutoML Tables
Why it fails: AutoML Tables is for structured tabular data, not images.

Full breakdown with real-world context →

Question 16hardmultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A healthcare company needs to extract structured data from scanned patient intake forms, including handwritten fields like patient name and date of birth. They require high accuracy and want to minimize custom code. Which Google Cloud solution should they use?

Trap 1: Vision API OCR

Vision API OCR can extract text from images, including handwriting, but it returns raw text without structure. It cannot directly map extracted text to specific fields like patient name or date of birth. Additional custom code would be needed to parse and structure the output, which contradicts the low-code requirement. Therefore, it is not the optimal solution for this scenario.

Trap 2: Document AI Form Parser

Document AI Form Parser is designed for general form parsing but may not handle handwritten text with high accuracy. It is optimized for printed text and common form fields. For handwritten fields, a specialized solution like Document AI Custom Extractor trained on handwriting samples would be more appropriate. Thus, Form Parser alone may not meet the accuracy requirement for handwritten patient data.

Trap 3: AutoML Vision

AutoML Vision is for image classification and object detection, not for extracting text from documents. It cannot perform OCR or understand form fields. While it could be trained to recognize form types, it would not extract the required data. Thus, it is unsuitable for this task, which requires document understanding and field extraction.

Study all Architecting Low-Code ML Solutions common traps →

A
Vision API OCR
Why it fails: Vision API OCR can extract text from images, including handwriting, but it returns raw text without structure. It cannot directly map extracted text to specific fields like patient name or date of birth. Additional custom code would be needed to parse and structure the output, which contradicts the low-code requirement. Therefore, it is not the optimal solution for this scenario.
B
Document AI Form Parser
Why it fails: Document AI Form Parser is designed for general form parsing but may not handle handwritten text with high accuracy. It is optimized for printed text and common form fields. For handwritten fields, a specialized solution like Document AI Custom Extractor trained on handwriting samples would be more appropriate. Thus, Form Parser alone may not meet the accuracy requirement for handwritten patient data.
C
Document AI Custom Extractor
Document AI Custom Extractor allows training a custom model on your own labeled documents, including handwritten fields. It can learn to recognize specific entities like patient name and date of birth from scanned forms. With minimal coding, you can label samples and train a model that achieves high accuracy. It is the best choice for extracting structured data from handwritten forms in a low-code manner.
D
AutoML Vision
Why it fails: AutoML Vision is for image classification and object detection, not for extracting text from documents. It cannot perform OCR or understand form fields. While it could be trained to recognize form types, it would not extract the required data. Thus, it is unsuitable for this task, which requires document understanding and field extraction.

Full breakdown with real-world context →

Question 17mediummulti select

Read the full Architecting Low-Code ML Solutions explanation →

A marketing team wants to build a model to predict customer lifetime value (CLV) using tabular data in BigQuery. They have limited ML expertise and want to use a low-code approach. They also need to explain which features contribute most to predictions for business stakeholders. Which TWO approaches should they use? (Choose two.)

Trap 1: Use BigQuery ML to train a deep neural network (DNN) model and use…

While BigQuery ML DNN is low-code, it does not provide built-in feature attributions like ML.EXPLAIN_PREDICT does for certain model types. The ML.PREDICT function only returns predictions, not explanations. Without additional tools, it would be difficult to explain feature contributions to stakeholders. Thus, it does not satisfy the requirement for model interpretability in this scenario.

Trap 2: Use Dataflow to preprocess data and train a custom TensorFlow model…

This approach requires significant coding and ML expertise, contradicting the low-code requirement. Dataflow and custom TensorFlow development are not low-code and would take considerable effort. While SHAP provides explanations, the overall solution is not aligned with the team's limited ML expertise. Therefore, it is not a suitable approach for this scenario.

Trap 3: Use BigQuery ML to train a boosted tree model and use the…

ML.FEATURE_IMPORTANCE provides global feature importance for boosted tree models, but it does not explain individual predictions. Stakeholders often need to understand why a specific customer received a particular CLV score, which requires local explanations. While the model is low-code, the lack of per-prediction explanations makes it less suitable for the requirement to explain predictions to stakeholders.

Study all Architecting Low-Code ML Solutions common traps →

A
Use BigQuery ML to train a deep neural network (DNN) model and use the ML.PREDICT function to get predictions.
Why it fails: While BigQuery ML DNN is low-code, it does not provide built-in feature attributions like ML.EXPLAIN_PREDICT does for certain model types. The ML.PREDICT function only returns predictions, not explanations. Without additional tools, it would be difficult to explain feature contributions to stakeholders. Thus, it does not satisfy the requirement for model interpretability in this scenario.
B
Use Dataflow to preprocess data and train a custom TensorFlow model with SHAP for explanations.
Why it fails: This approach requires significant coding and ML expertise, contradicting the low-code requirement. Dataflow and custom TensorFlow development are not low-code and would take considerable effort. While SHAP provides explanations, the overall solution is not aligned with the team's limited ML expertise. Therefore, it is not a suitable approach for this scenario.
C
Use BigQuery ML to train a linear regression model and use the ML.EXPLAIN_PREDICT function to get feature attributions.
BigQuery ML's linear regression is a low-code, SQL-based approach suitable for tabular data. The ML.EXPLAIN_PREDICT function provides Shapley-based explanations that show how each feature contributed to a prediction, which is essential for explaining model outputs to stakeholders. This combination meets both the low-code requirement and the need for interpretability, making it a strong choice for CLV prediction.
D
Use Vertex AI AutoML Tables to train a model and rely on its built-in feature importance scores.
Vertex AI AutoML Tables is a low-code solution that automatically builds and tunes models on tabular data. It provides feature importance scores that indicate which features were most influential globally, which can be shared with stakeholders. While it requires exporting data to Vertex AI, it still minimizes coding and offers interpretability, making it a valid alternative for CLV prediction when combined with the need for explanations.
E
Use BigQuery ML to train a boosted tree model and use the ML.FEATURE_IMPORTANCE function to get global feature importance.
Why it fails: ML.FEATURE_IMPORTANCE provides global feature importance for boosted tree models, but it does not explain individual predictions. Stakeholders often need to understand why a specific customer received a particular CLV score, which requires local explanations. While the model is low-code, the lack of per-prediction explanations makes it less suitable for the requirement to explain predictions to stakeholders.

Full breakdown with real-world context →

Question 18mediummultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A retail company wants to forecast daily sales for the next 90 days using historical sales data. They have no ML expertise and want a fully managed service that automatically handles time series modeling, including seasonality and holidays. Which Google Cloud service should they use?

Trap 1: AI Platform Prediction

AI Platform Prediction is a service for deploying and serving trained models, not for training or forecasting. It requires a pre-trained model and does not provide any time series modeling capabilities. The retail company would need to build and train a model elsewhere, which contradicts the need for a fully managed, no-code forecasting solution.

Trap 2: BigQuery ML with ARIMA_PLUS

BigQuery ML with ARIMA_PLUS is a powerful time series model, but it requires writing SQL to create and train the model. It does handle seasonality and holidays, but the retail company lacks ML expertise and may not be comfortable with SQL. Additionally, while it is low-code, it is not a fully managed service with a UI for non-technical users, so it does not meet the requirement for a no-code solution.

Trap 3: AutoML Tables

AutoML Tables is designed for general tabular regression and classification tasks, not specifically for time series forecasting. It does not automatically handle temporal dependencies, seasonality, or holidays unless the user manually engineers lag features and date-related features. For forecasting daily sales, this manual effort is error-prone and not low-code, so it is not the best fit.

Study all Architecting Low-Code ML Solutions common traps →

A
AI Platform Prediction
Why it fails: AI Platform Prediction is a service for deploying and serving trained models, not for training or forecasting. It requires a pre-trained model and does not provide any time series modeling capabilities. The retail company would need to build and train a model elsewhere, which contradicts the need for a fully managed, no-code forecasting solution.
B
BigQuery ML with ARIMA_PLUS
Why it fails: BigQuery ML with ARIMA_PLUS is a powerful time series model, but it requires writing SQL to create and train the model. It does handle seasonality and holidays, but the retail company lacks ML expertise and may not be comfortable with SQL. Additionally, while it is low-code, it is not a fully managed service with a UI for non-technical users, so it does not meet the requirement for a no-code solution.
C
Vertex AI Forecast
Vertex AI Forecast is a fully managed, no-code service specifically designed for time series forecasting. It automatically handles seasonality, holidays, and missing data, and provides a user-friendly interface for uploading data and generating forecasts. The retail company can use it without ML expertise, making it the ideal choice for forecasting daily sales.
D
AutoML Tables
Why it fails: AutoML Tables is designed for general tabular regression and classification tasks, not specifically for time series forecasting. It does not automatically handle temporal dependencies, seasonality, or holidays unless the user manually engineers lag features and date-related features. For forecasting daily sales, this manual effort is error-prone and not low-code, so it is not the best fit.

Full breakdown with real-world context →

Question 19mediummultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A retail company wants to build a model to predict customer lifetime value (CLV) using tabular data with 50,000 rows and 20 features. They have limited ML expertise and want to minimize coding. They also need to be able to explain which features most influence the predictions. Which Google Cloud service should they use?

Trap 1: BigQuery ML with a linear regression model.

BigQuery ML is low-code and can train a linear regression model via SQL, but it may not provide the same level of automatic feature importance and explainability as AutoML Tables. Additionally, linear regression assumes a linear relationship, which may not capture complex patterns in CLV prediction. The company wants to minimize coding, but BigQuery ML still requires SQL knowledge and manual feature engineering.

Trap 2: Cloud TPU with a custom TensorFlow model.

Cloud TPUs are specialized hardware for training deep learning models, typically used with TensorFlow. This approach requires significant ML expertise and coding to build and train a custom model. It is not low-code and does not provide built-in explainability. The dataset size (50,000 rows) does not justify TPU usage, which is overkill and adds unnecessary complexity.

Trap 3: AI Platform Prediction with a pre-built scikit-learn model.

AI Platform Prediction (now part of Vertex AI) is for deploying and serving models, not for training. Using a pre-built scikit-learn model still requires coding to train and tune the model. It does not offer automated feature engineering or built-in explainability. This option does not meet the low-code requirement and lacks the automation of AutoML Tables.

Study all Architecting Low-Code ML Solutions common traps →

A
BigQuery ML with a linear regression model.
Why it fails: BigQuery ML is low-code and can train a linear regression model via SQL, but it may not provide the same level of automatic feature importance and explainability as AutoML Tables. Additionally, linear regression assumes a linear relationship, which may not capture complex patterns in CLV prediction. The company wants to minimize coding, but BigQuery ML still requires SQL knowledge and manual feature engineering.
B
Vertex AI AutoML Tables (now Vertex AI Tabular).
Vertex AI AutoML Tables automates feature engineering, model selection, and hyperparameter tuning, requiring minimal coding. It provides feature importance and explainable AI capabilities, showing which features influence predictions. It is designed for tabular data and can handle datasets of this size. This aligns with the company's need for low-code and explainability.
C
Cloud TPU with a custom TensorFlow model.
Why it fails: Cloud TPUs are specialized hardware for training deep learning models, typically used with TensorFlow. This approach requires significant ML expertise and coding to build and train a custom model. It is not low-code and does not provide built-in explainability. The dataset size (50,000 rows) does not justify TPU usage, which is overkill and adds unnecessary complexity.
D
AI Platform Prediction with a pre-built scikit-learn model.
Why it fails: AI Platform Prediction (now part of Vertex AI) is for deploying and serving models, not for training. Using a pre-built scikit-learn model still requires coding to train and tune the model. It does not offer automated feature engineering or built-in explainability. This option does not meet the low-code requirement and lacks the automation of AutoML Tables.

Full breakdown with real-world context →

Question 20hardmultiple choice

Read the full Architecting Low-Code ML Solutions explanation →

A healthcare startup wants to build a model to predict patient readmission risk using structured electronic health records (EHR) data. They have a small dataset of 5,000 records and need to ensure the model is interpretable for regulatory compliance. They prefer a low-code solution but require feature importance explanations. Which approach should they take?

Trap 1: Use AutoML Tables with export to BigQuery and then apply LIME

Exporting an AutoML Tables model to BigQuery and applying LIME is a convoluted approach. AutoML Tables models can be exported to BigQuery ML, but LIME is not natively supported in BigQuery ML. Implementing LIME would require additional coding and infrastructure. This approach is not low-code and adds unnecessary complexity, making it unsuitable for the startup's needs.

Trap 2: Use Vertex AI AutoML Tables and enable feature importance

Vertex AI AutoML Tables can provide feature importance, but it requires a minimum of 1,000 rows per class for classification, and with 5,000 records, it may be sufficient if the classes are balanced. However, AutoML Tables is not ideal for small datasets because it may overfit, and the feature importance is global, not per-instance. For regulatory compliance, per-instance explanations are often needed, which AutoML Tables does not provide natively.

Trap 3: Use Vertex AI Training with a custom TensorFlow model and integrate…

Using Vertex AI Training with a custom TensorFlow model and SHAP requires significant coding and ML expertise, which contradicts the low-code preference. While SHAP provides excellent interpretability, it adds complexity and is not a low-code solution. The startup would need to manage training infrastructure, write code, and integrate SHAP, which is overkill for a small dataset and not aligned with their requirements.

Study all Architecting Low-Code ML Solutions common traps →

A
Use BigQuery ML with a logistic regression model and use the ML.EXPLAIN_PREDICT function
BigQuery ML supports logistic regression for binary classification and provides ML.EXPLAIN_PREDICT, which returns per-instance feature attributions. This is crucial for regulatory compliance and interpretability. With 5,000 records, logistic regression is less prone to overfitting than complex models. The startup can use SQL, which is low-code compared to Python, and get detailed explanations for each prediction.
B
Use AutoML Tables with export to BigQuery and then apply LIME
Why it fails: Exporting an AutoML Tables model to BigQuery and applying LIME is a convoluted approach. AutoML Tables models can be exported to BigQuery ML, but LIME is not natively supported in BigQuery ML. Implementing LIME would require additional coding and infrastructure. This approach is not low-code and adds unnecessary complexity, making it unsuitable for the startup's needs.
C
Use Vertex AI AutoML Tables and enable feature importance
Why it fails: Vertex AI AutoML Tables can provide feature importance, but it requires a minimum of 1,000 rows per class for classification, and with 5,000 records, it may be sufficient if the classes are balanced. However, AutoML Tables is not ideal for small datasets because it may overfit, and the feature importance is global, not per-instance. For regulatory compliance, per-instance explanations are often needed, which AutoML Tables does not provide natively.
D
Use Vertex AI Training with a custom TensorFlow model and integrate SHAP
Why it fails: Using Vertex AI Training with a custom TensorFlow model and SHAP requires significant coding and ML expertise, which contradicts the low-code preference. While SHAP provides excellent interpretability, it adds complexity and is not a low-code solution. The startup would need to manage training infrastructure, write code, and integrate SHAP, which is overkill for a small dataset and not aligned with their requirements.

Full breakdown with real-world context →

Continue with 20-question session →

Free account

Track your progress over time

Create a free account to save your results and see which topics improve across sessions.

Focused Architecting Low-Code ML Solutions sessions

Start a Architecting Low-Code ML Solutions only practice session

Every question in these sessions is drawn from the Architecting Low-Code ML Solutions domain — nothing else.

10 questions 20 questions 30 questions 50 questions

Browse all Architecting Low-Code ML Solutions questions →Mixed PMLE session

Frequently asked questions

What does the PMLE exam test about Architecting Low-Code ML Solutions?: Be able to pick the right low-code service: BigQuery ML model type by prediction task, pretrained API by data modality, and human review for high-risk cases. The most important thing is matching the service to the required output and risk level.
How should I use these practice questions?: Select your answer before revealing the explanation. Then read why each option is right or wrong — this active recall approach builds retention far faster than re-reading notes.
Can I practise just Architecting Low-Code ML Solutions questions in a focused session?: Yes — the session launcher on this page draws every question from the Architecting Low-Code ML Solutions domain. Use a 10-question session first to gauge your baseline, then move to 20 or 30 once the weak spots are clear.
Where can I practise other PMLE topics?: Use the topic links above to move to related areas, or go back to the PMLE question bank to see all topics.
Are these real exam questions or dumps?: These are original practice questions written to test the same concepts the PMLE exam covers. They are not copied from any real exam or dump site.

Architecting Low-Code ML Solutions only

10 questions 20 questions 30 questions 50 questions

Mixed PMLE session

Track your progress

A free account saves results across sessions and highlights which topics need work.

Study resources

All PMLE questions Architecting Low-Code ML Solutions domain overview PMLE exam guide

Exam traps to avoid

▸Assuming BigQuery ML requires exporting data or writing Python; it trains models directly with SQL CREATE MODEL statements
▸Confusing pretrained API use cases: Natural Language for sentiment versus Document AI for PDF form and table extraction
▸Forgetting that high-risk loan or financial decisions need a human review step, not fully automated API output

Architecting Low-Code ML Solutions practice questions

What to know about Architecting Low-Code ML Solutions

Common Architecting Low-Code ML Solutions exam traps

Architecting Low-Code ML Solutions questions

A company wants to use BigQuery ML to train a DNN_CLASSIFIER model on a dataset with 100 million rows. They are concerned about training time and cost. Which approach can help optimize training performance while staying within BigQuery ML?

A company needs to analyze customer feedback from app reviews to identify common themes and sentiment. They have millions of reviews in multiple languages. Which combination of pre-built APIs should they use?

A retail company wants to build a recommendation system for their e-commerce website. They have user purchase history and product metadata. Which Google Cloud service is most suitable for building a 'frequently bought together' recommendation model with minimal custom ML development?

A retailer wants to implement a recommendation engine that suggests products based on a user's current cart. They have limited ML expertise and want a quick deployment. Which Recommendations AI model type should they use?

A company needs to detect objects in real-time from a live video feed. They want to use a pre-trained model with minimal setup. Which Google Cloud service should they use?

A data scientist wants to use BigQuery ML for time-series forecasting. They need to evaluate model accuracy and compare different models. Which TWO BigQuery ML functions should they use?

A company needs to classify images of products into categories (e.g., electronics, clothing, food). They have labeled images and want to use a low-code solution on Google Cloud. Which service is suitable for this task?

A company wants to transcribe customer service calls in real-time. The audio is telephony quality (8 kHz). Which Speech-to-Text model should they specify?

A company uses BigQuery ML with a remote model calling Vertex AI's pre-trained image classification model. They need to classify images stored in Cloud Storage buckets. What is the correct approach?

A data scientist wants to use AutoML Tables to build a binary classification model for loan default prediction. The dataset has 200 features and 1 million rows, with highly imbalanced classes. Which TWO options should they consider? (Choose 2)

A company needs to detect objects in live video streams from security cameras. They require low-latency predictions and want to minimise operational overhead. Which TWO services should they use? (Choose 2)

A company wants to use Document AI to process a large volume of invoices. They need to extract line items and also have a human review the extracted data for accuracy. Which THREE features should they use? (Choose 3)

A data engineer is using BigQuery ML with a BOOSTED_TREE_CLASSIFIER model. After training, they want to evaluate the model and understand which features contribute most to predictions. Which THREE BigQuery ML functions should they use?

A company needs to build a custom model to classify images of products into categories. They have a large labeled dataset. They want to use AutoML but are unsure which options support image classification. Which TWO AutoML products support image classification?

A healthcare company needs to extract structured data from scanned patient intake forms, including handwritten fields like patient name and date of birth. They require high accuracy and want to minimize custom code. Which Google Cloud solution should they use?

A retail company wants to forecast daily sales for the next 90 days using historical sales data. They have no ML expertise and want a fully managed service that automatically handles time series modeling, including seasonality and holidays. Which Google Cloud service should they use?

Track your progress over time

Start a Architecting Low-Code ML Solutions only practice session

Related PMLE topic practice pages

Automating and Orchestrating ML Pipelines practice questions

Serving and Scaling Models practice questions