Google Professional Machine Learning Engineer Exam Dumps and Practice Test Questions Set 10 Q136-150

Visit here for our full Google Professional Machine Learning Engineer exam dumps and practice test questions.

Question 136

A bank wants to implement real-time credit scoring for loan applications. The system must ingest transaction data, credit history, and demographic information continuously, provide low-latency scoring, and adapt to changing risk patterns. Which architecture is most appropriate?

A) Batch process loan applications nightly and manually evaluate credit risk
B) Use Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online scoring
C) Store transaction and demographic data in spreadsheets and manually calculate credit scores
D) Train a model once on historical applications and deploy it permanently

Answer: B

Explanation:

Batch processing loan applications nightly and manually evaluating credit risk is inadequate for real-time credit scoring. Loan decisions must often be made instantly to meet customer expectations. Nightly batch processing introduces delays, meaning applicants may wait hours or days for decisions, which can reduce business competitiveness and customer satisfaction. Manual evaluation is slow, error-prone, and does not scale for large numbers of applications. Furthermore, batch workflows do not allow models to adapt continuously to changing customer behavior, new credit patterns, or regulatory requirements, limiting predictive accuracy. This approach is unsuitable for modern, real-time banking operations.

Using Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online scoring is the most appropriate architecture. Pub/Sub ingests transaction data, credit history, and demographic information continuously, ensuring no events are missed. Dataflow pipelines compute derived features in real time, such as credit utilization ratios, spending patterns, and historical repayment trends, which are critical for accurate scoring. Vertex AI Prediction serves low-latency credit scores, enabling immediate loan approval or rejection decisions. Continuous retraining pipelines allow the model to adapt to evolving customer behavior, new financial products, or updated regulatory requirements. Autoscaling ensures the system handles peak application volumes efficiently. Logging, monitoring, and reproducibility provide operational reliability, auditability, and compliance. This architecture enables scalable, low-latency, continuously adaptive credit scoring.

Storing transaction and demographic data in spreadsheets and manually calculating credit scores is impractical. Spreadsheets cannot process high-volume, high-frequency application data or integrate multiple sources effectively. Manual computation is slow, error-prone, and non-reproducible. This approach cannot scale to enterprise-level credit scoring or provide low-latency responses, making it unsuitable for production banking environments.

Training a model once on historical applications and deploying it permanently is inadequate. Credit patterns, customer behavior, and economic conditions change over time. A static model cannot adapt, resulting in inaccurate scores and potential financial risk. Continuous retraining and real-time feature computation are necessary to maintain accuracy and compliance.

The optimal solution is Pub/Sub for real-time ingestion, Dataflow for feature computation, and Vertex AI Prediction for online scoring, providing scalable, low-latency, and continuously adaptive credit scoring.

Question 137

A healthcare provider wants to predict patient hospital stay duration using EHR data, lab results, and imaging. The model must scale to large datasets, ensure reproducibility, comply with privacy regulations, and adapt as new patient data becomes available. Which approach is most appropriate?

A) Download all patient data locally and train models manually
B) Use BigQuery for structured data, Cloud Storage for unstructured data, and Vertex AI Pipelines for preprocessing, training, and deployment
C) Store data in spreadsheets and manually estimate stay duration
D) Train a model once using sample data and deploy it permanently

Answer: B

Explanation:

Downloading all patient data locally and training models manually is unsuitable due to privacy, compliance, and scalability concerns. Healthcare data is sensitive, subject to HIPAA and other regulations, and local storage increases the risk of unauthorized access. Manual training workflows are slow, error-prone, and non-reproducible. They cannot handle large, heterogeneous datasets combining structured EHR data, lab results, and imaging efficiently. Preprocessing data manually is inconsistent and does not allow automated retraining, which is essential to maintain predictive accuracy as new patient data arrives. This approach is inadequate for production healthcare predictive systems.

Using BigQuery for structured data, Cloud Storage for unstructured data, and Vertex AI Pipelines for preprocessing, training, and deployment is the most appropriate solution. BigQuery efficiently manages structured EHR datasets, enabling scalable querying and aggregation. Cloud Storage stores clinical notes and imaging data, providing secure, scalable access. Vertex AI Pipelines orchestrate preprocessing, feature extraction, training, and deployment reproducibly, ensuring consistent processing across diverse data types. Continuous retraining pipelines allow the model to adapt to new patient data, improving accuracy over time. Logging, monitoring, and experiment tracking ensure operational reliability, reproducibility, and compliance with privacy regulations. Autoscaling supports the processing of large datasets without latency degradation. This architecture provides secure, scalable, reproducible, and continuously adaptive predictions for patient hospital stay duration.

Storing data in spreadsheets and manually estimating stay duration is impractical. Spreadsheets cannot process large-scale structured and unstructured data efficiently. Manual estimation is slow, error-prone, non-reproducible, and does not support continuous retraining or low-latency predictions.

Training a model once using sample data and deploying it permanently is inadequate. Patient populations, clinical procedures, and hospital conditions change over time. A static model cannot adapt to new patterns, resulting in reduced predictive accuracy. Continuous retraining and automated pipelines are essential for operational effectiveness.

The optimal solution is BigQuery, Cloud Storage, and Vertex AI Pipelines for secure, scalable, reproducible, and continuously adaptive patient stay predictions.

Question 138

A retailer wants to forecast inventory demand for thousands of products across multiple stores. The system must scale to millions of historical sales records, support feature reuse, and continuously update predictions as new sales and promotions occur. Which solution is most appropriate?

A) Train separate models locally for each store using spreadsheets
B) Use Vertex AI Feature Store for centralized features and Vertex AI Training for distributed forecasting
C) Store historical sales data in Cloud SQL and train a single global linear regression model
D) Use a simple rule-based system based on last year’s sales

Answer: B

Explanation:

Training separate models locally for each store using spreadsheets is impractical. Large-scale retail datasets involve millions of records and multiple feature types such as historical sales, promotions, holidays, and weather. Local training cannot efficiently process such volumes, is slow, error-prone, and non-reproducible. Manual feature management leads to redundancy and inconsistency. Automated retraining pipelines are difficult to implement locally, and scalability is limited. This approach is unsuitable for enterprise demand forecasting.

Using Vertex AI Feature Store for centralized features and Vertex AI Training for distributed forecasting is the most appropriate solution. Feature Store ensures consistent, reusable feature definitions across multiple models, reducing duplication and guaranteeing consistency between training and serving. Vertex AI Training supports distributed training across GPUs or TPUs, efficiently handling millions of records while learning complex patterns in sales, promotions, and external factors. Pipelines automate feature updates, retraining, and model versioning, ensuring forecasts continuously improve as new sales and promotion data arrive. Autoscaling allows the system to handle high data volumes efficiently. Logging, monitoring, and experiment tracking ensure reproducibility, operational reliability, and governance compliance. This architecture supports scalable, accurate, and continuously updated inventory demand forecasting across thousands of products and multiple stores.

Storing historical sales data in Cloud SQL and training a single global linear regression model is inadequate. Cloud SQL is not optimized for large-scale analytical workloads, and linear regression cannot capture complex nonlinear relationships. A single global model may underfit, producing inaccurate forecasts and lacking support for localized or store-specific patterns.

Using a simple rule-based system based on last year’s sales is insufficient. Rule-based approaches cannot account for promotions, holidays, weather, or changing sales trends. They lack scalability, automation, and predictive accuracy, making them unsuitable for enterprise forecasting.

The optimal solution is Vertex AI Feature Store combined with Vertex AI Training, providing scalable, reusable, and continuously updated demand forecasts across multiple products and stores.

Question 139

A logistics company wants to optimize delivery routes in real time using vehicle telemetry, traffic data, and weather information. The system must scale to thousands of vehicles, provide low-latency predictions, and continuously adapt to changing conditions. Which architecture is most appropriate?

A) Batch process delivery and traffic data daily, and manually update routes
B) Use Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online routing optimization
C) Store vehicle and traffic data in spreadsheets and manually compute optimal routes
D) Train a route optimization model once and deploy it permanently

Answer: B

Explanation:

Batch processing, delivering, and traffic data daily,y and manually updating routes daily is inadequate for real-time delivery optimization. Traffic and weather conditions change rapidly, and batch updates introduce delays that result in outdated route recommendations. Manual computation cannot scale to thousands of vehicles and is prone to human error. Without continuous retraining and real-time processing, the system cannot adapt to changing conditions, which reduces operational efficiency and customer satisfaction. This approach is unsuitable for enterprise-scale logistics.

Using Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online routing optimization is the most appropriate solution. Pub/Sub ingests vehicle telemetry, traffic, and weather data continuously, ensuring no events are missed. Dataflow pipelines compute features such as estimated route delays, congestion impact, and weather influence on travel time. Vertex AI Prediction serves low-latency routing decisions to dispatch systems, enabling real-time adjustments. Continuous retraining pipelines ensure models adapt to evolving traffic patterns, vehicle behavior, and environmental changes. Autoscaling ensures high-throughput processing during peak periods. Logging, monitoring, and reproducibility provide operational reliability, traceability, and compliance. This architecture ensures scalable, low-latency, and continuously adaptive delivery route optimization.

Storing vehicle and traffic data in spreadsheets and manually computing optimal routes is impractical. Spreadsheets cannot handle high-frequency telemetry data, traffic updates, or weather information efficiently. Manual calculations are slow, error-prone, non-reproducible, and unsuitable for real-time operations.

Training a route optimization model once and deploying it permanently is insufficient. Traffic, vehicle availability, and environmental conditions evolve constantly, and a static model cannot adapt, resulting in suboptimal route recommendations. Continuous retraining and real-time data ingestion are essential for accurate, operationally effective route optimization.

The optimal solution is Pub/Sub for real-time ingestion, Dataflow for feature computation, and Vertex AI Prediction for online routing optimization, providing scalable, low-latency, and continuously adaptive delivery routes.

Question 140

A telecommunications company wants to predict network outages using logs from multiple devices in real time. The system must scale to handle millions of log entries per second, provide low-latency anomaly detection, and allow continuous retraining. Which solution is most appropriate?

A) Batch process logs nightly and manually inspects for anomalies
B) Use Pub/Sub for log ingestion, Dataflow for feature computation, and Vertex AI Prediction for online anomaly detection
C) Store logs in spreadsheets and manually compute anomalies
D) Train a model once on historical logs and deploy it permanently

Answer: B

Explanation:

Batch processing logs nightly and manually inspecting for anomalies is insufficient for real-time network outage prediction. Network conditions can deteriorate rapidly, and batch processing introduces delays, leaving potential outages undetected. Manual inspection is slow, error-prone, and does not scale to millions of log entries per second. Without continuous retraining, models cannot adapt to new network behaviors or evolving device patterns. This approach cannot meet operational requirements for low-latency detection and proactive outage prevention.

Using Pub/Sub for log ingestion, Dataflow for feature computation, and Vertex AI Prediction for online anomaly detection is the most suitable solution. Pub/Sub ingests high-throughput log data in real time from multiple devices, ensuring all events are captured. Dataflow pipelines compute features such as error rates, latency spikes, packet loss, and device-specific metrics, providing meaningful inputs for anomaly detection models. Vertex AI Prediction delivers low-latency scoring to detect anomalies immediately, enabling automated alerts and mitigation. Continuous retraining pipelines incorporate new log patterns to improve model accuracy over time. Autoscaling handles spikes in log volume efficiently. Logging, monitoring, and reproducibility provide operational reliability, auditability, and regulatory compliance. This architecture ensures scalable, low-latency, and continuously adaptive network anomaly detection.

Storing logs in spreadsheets and manually computing anomalies is impractical. Spreadsheets cannot process millions of log entries efficiently. Manual computation is slow, error-prone, non-reproducible, and unsuitable for real-time anomaly detection.

Training a model once on historical logs and deploying it permanently is inadequate. Network patterns and device behavior evolve constantly. Static models cannot adapt to new anomalies, reducing detection accuracy. Continuous retraining and real-time feature computation are essential to maintain operational effectiveness.

The optimal solution is Pub/Sub for ingestion, Dataflow for feature computation, and Vertex AI Prediction for online anomaly detection, ensuring scalable, low-latency, and continuously adaptive outage prediction.

Question 141

A retailer wants to forecast daily product demand using historical sales, promotions, holidays, and weather data. The system must scale to millions of records, allow feature reuse, and continuously update forecasts. Which solution is most appropriate?

Answer: B

Explanation:

Training separate models locally for each store using spreadsheets is impractical. Millions of records across multiple stores and product types cannot be efficiently handled with local spreadsheets. Manual training is slow, error-prone, and non-reproducible. Managing features manually leads to redundancy, inconsistency, and difficulty scaling. Automated retraining pipelines are hard to implement locally, and feature reuse is limited. This approach is unsuitable for enterprise-scale demand forecasting.

Using Vertex AI Feature Store for centralized features and Vertex AI Training for distributed forecasting is the most appropriate solution. Feature Store ensures consistent, reusable features across multiple models, reducing duplication and ensuring consistency between training and serving. Vertex AI Training supports distributed model training across GPUs or TPUs, efficiently processing millions of historical records while capturing complex patterns in sales, promotions, holidays, and weather. Automated pipelines manage feature updates, retraining, and model versioning, ensuring forecasts continuously improve as new sales and promotion data becomes available. Autoscaling allows efficient handling of high data volumes. Logging, monitoring, and experiment tracking provide reproducibility, operational reliability, and governance compliance. This architecture enables scalable, accurate, and continuously updated demand forecasts across multiple products and stores.

Storing historical sales data in Cloud SQL and training a single global linear regression model is insufficient. Cloud SQL is not optimized for large-scale analytical workloads, and linear regression cannot capture complex, non-linear relationships. A single model may underfit and produce inaccurate forecasts. Feature reuse is limited, and localized store-level patterns may be missed.

Using a simple rule-based system based on last year’s sales is inadequate. Rule-based approaches cannot account for promotions, holidays, or changing trends. They lack scalability, automation, and predictive accuracy, making them unsuitable for enterprise demand forecasting.

The optimal solution is Vertex AI Feature Store combined with Vertex AI Training, providing scalable, reusable, and continuously updated product demand forecasts across multiple stores.

Question 142

A bank wants to implement real-time fraud detection for online transactions. The system must handle high transaction volumes, provide low-latency scoring, and continuously adapt to new fraud patterns. Which architecture is most appropriate?

A) Batch process transactions daily and manually review suspicious activity
B) Use Pub/Sub for transaction ingestion, Dataflow for feature engineering, and Vertex AI Prediction for online fraud scoring
C) Store transactions in spreadsheets and manually compute fraud risk
D) Train a fraud detection model once per year and deploy it permanently

Answer: B

Explanation:

Batch processing transactions daily and manually reviewing suspicious activity is insufficient for real-time fraud detection. Fraudulent transactions can occur within seconds, and daily batch processing introduces unacceptable delays, leaving potential fraud undetected. Manual review cannot scale to thousands or millions of transactions per second, is prone to human error, and does not provide low-latency detection. Furthermore, batch workflows cannot continuously adapt to evolving fraud patterns, reducing overall accuracy and effectiveness. This approach is unsuitable for modern, real-time financial operations.

Using Pub/Sub for transaction ingestion, Dataflow for feature engineering, and Vertex AI Prediction for online fraud scoring is the most appropriate solution. Pub/Sub ensures high-throughput ingestion of online transactions in real time. Dataflow pipelines compute derived features such as transaction frequency, spending patterns, location anomalies, and device behavior, which are critical inputs for fraud detection models. Vertex AI Prediction delivers low-latency scoring, enabling immediate alerts and interventions. Continuous retraining pipelines allow models to incorporate new fraud patterns, improve detection accuracy, and adapt to evolving financial behaviors. Autoscaling ensures the system can handle spikes in transaction volume without latency degradation. Logging, monitoring, and reproducibility provide operational reliability, auditability, and regulatory compliance. This architecture supports scalable, low-latency, and continuously adaptive fraud detection for online banking.

Storing transactions in spreadsheets and manually computing fraud risk is impractical. Spreadsheets cannot manage high-frequency, high-volume transaction data. Manual computation is slow, error-prone, non-reproducible, and cannot support continuous retraining. This approach is unsuitable for operational fraud detection at scale.

Training a fraud detection model once per year and deploying it permanently is inadequate. Fraud patterns evolve quickly, and a static model cannot detect emerging tactics. Annual retraining is too infrequent, resulting in decreased accuracy and increased risk. Real-time adaptation and continuous retraining are essential for reliable fraud detection.

The optimal solution is Pub/Sub for real-time ingestion, Dataflow for feature engineering, and Vertex AI Prediction for online fraud scoring, providing scalable, low-latency, and continuously adaptive detection.

Question 143

A logistics company wants to forecast estimated delivery times using vehicle telemetry, traffic data, and weather conditions. Predictions must be low-latency, scalable, and continuously adaptive. Which solution is most appropriate?

A) Batch process delivery data daily and manually update predictions
B) Use Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online forecasting
C) Store delivery data in spreadsheets and manually estimate delivery times
D) Train a model once on historical data and deploy it permanently

Answer: B

Explanation:

Batch processing delivery data daily and manually updating predictions is inadequate for real-time logistics forecasting. Delivery times are influenced by dynamic variables such as traffic congestion, vehicle speed, and weather conditions, which change continuously. Nightly batch processing introduces delays, making forecasts outdated and operationally ineffective. Manual updates cannot scale to thousands of vehicles and are prone to human error. Without continuous retraining and real-time feature computation, predictions will degrade in accuracy over time, limiting operational effectiveness.

Using Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online forecasting is the most suitable solution. Pub/Sub continuously ingests telemetry from vehicles, traffic updates, and weather information. Dataflow pipelines compute derived features such as congestion impact, route delays, and estimated speed. Vertex AI Prediction provides low-latency forecasts to operational systems, allowing immediate adjustments to routes and delivery schedules. Continuous retraining pipelines ensure models adapt to evolving patterns in traffic, vehicle behavior, and environmental conditions. Autoscaling handles peak delivery volumes efficiently. Logging, monitoring, and reproducibility provide operational reliability, traceability, and compliance. This architecture supports scalable, low-latency, continuously adaptive delivery time predictions.

Storing delivery data in spreadsheets and manually estimating delivery times is impractical. Spreadsheets cannot process high-volume, high-frequency telemetry and traffic data. Manual estimation is slow, error-prone, non-reproducible, and cannot support continuous retraining, making it unsuitable for real-time logistics operations.

Training a model once on historical data and deploying it permanently is insufficient. Delivery patterns change due to traffic, weather, and operational variations. Static models cannot adapt, resulting in inaccurate predictions. Continuous retraining and online feature computation are necessary to maintain accuracy.

The optimal solution is Pub/Sub for real-time ingestion, Dataflow for feature computation, and Vertex AI Prediction for online forecasting, providing scalable, low-latency, and continuously adaptive delivery predictions.

Question 144

A retailer wants to forecast inventory demand across multiple stores using historical sales, promotions, holidays, and weather. The system must scale to millions of records, support feature reuse, and continuously update forecasts. Which solution is most appropriate?

Answer: B

Explanation:

Training separate models locally for each store using spreadsheets is impractical. Retail datasets consist of millions of records and multiple feature types including sales history, promotions, holidays, and weather. Local spreadsheets cannot handle this scale efficiently. Manual training is slow, error-prone, and non-reproducible. Managing features separately for each store introduces redundancy and inconsistency. Automated retraining pipelines are difficult to implement locally, and feature reuse is limited. This approach is unsuitable for enterprise-level demand forecasting.

Using Vertex AI Feature Store for centralized features and Vertex AI Training for distributed forecasting is the most appropriate solution. Feature Store ensures consistent, reusable features across multiple models, reducing redundancy and ensuring consistency between training and serving. Vertex AI Training supports distributed training across GPUs or TPUs, efficiently processing millions of historical records while capturing complex patterns across promotions, holidays, and weather. Pipelines automate feature updates, retraining, and versioning, ensuring forecasts continuously improve as new sales and promotion data becomes available. Autoscaling allows efficient handling of large data volumes. Logging, monitoring, and experiment tracking provide reproducibility, operational reliability, and governance compliance. This architecture enables scalable, accurate, and continuously updated inventory demand forecasting across multiple products and stores.

Storing historical sales data in Cloud SQL and training a single global linear regression model is insufficient. Cloud SQL is not optimized for large-scale analytical workloads, and linear regression cannot capture complex, non-linear patterns. A single model may underfit and produce inaccurate forecasts, lacking localized store-level precision.

Using a simple rule-based system based on last year’s sales is inadequate. Rule-based approaches cannot account for changing trends, promotions, holidays, or weather. They lack scalability, automation, and predictive accuracy, making them unsuitable for enterprise-level forecasting.

The optimal solution is Vertex AI Feature Store combined with Vertex AI Training, providing scalable, reusable, and continuously updated inventory demand forecasts.

Question 145

A telecommunications company wants to detect network failures in real time using logs from thousands of devices. The system must scale to high log volumes, provide low-latency detection, and continuously adapt to new failure patterns. Which solution is most appropriate?

A) Batch process logs nightly and manually inspect for failures
B) Use Pub/Sub for log ingestion, Dataflow for feature computation, and Vertex AI Prediction for online anomaly detection
C) Store logs in spreadsheets and manually compute failure metrics
D) Train a model once on historical logs and deploy permanently

Answer: B

Explanation:

Batch processing logs nightly and manually inspecting for failures is inadequate for real-time network failure detection. Network failures can occur quickly, and delays introduced by nightly batch processing may prevent timely detection, leading to service outages and decreased reliability. Manual inspection cannot scale to thousands of devices generating millions of log entries per second. Additionally, batch workflows do not support continuous model retraining or adaptation to new failure patterns, limiting prediction accuracy. This approach is unsuitable for modern telecommunication networks requiring operational real-time monitoring.

Using Pub/Sub for log ingestion, Dataflow for feature computation, and Vertex AI Prediction for online anomaly detection is the most appropriate solution. Pub/Sub enables high-throughput real-time ingestion of device logs, ensuring no events are missed. Dataflow pipelines process logs continuously, extracting meaningful features such as error rates, packet loss, latency spikes, and correlations between device metrics. Vertex AI Prediction delivers low-latency anomaly detection, allowing immediate identification of potential failures. Continuous retraining pipelines allow models to adapt to new device behavior, evolving network conditions, and emerging failure patterns. Autoscaling ensures the system can handle peak traffic efficiently. Logging, monitoring, and reproducibility provide operational reliability, traceability, and regulatory compliance. This architecture supports scalable, low-latency, and continuously adaptive network failure detection.

Storing logs in spreadsheets and manually computing failure metrics is impractical. Spreadsheets cannot handle the high volume and velocity of device logs efficiently. Manual computation is slow, error-prone, and non-reproducible. This approach cannot provide real-time detection or continuous adaptation, making it unsuitable for operational network monitoring.

Training a model once on historical logs and deploying permanently is insufficient. Network conditions and device behaviors evolve, and a static model cannot detect new types of failures, reducing reliability. Continuous retraining and real-time data ingestion are necessary to maintain accurate, operationally effective predictions.

The optimal solution is Pub/Sub for log ingestion, Dataflow for feature computation, and Vertex AI Prediction for online anomaly detection, providing scalable, low-latency, and continuously adaptive detection.

Question 146

A healthcare provider wants to predict patient readmission risk using EHR data, lab results, and imaging. The model must comply with privacy regulations, scale with growing datasets, and allow reproducible training pipelines. Which approach is most appropriate?

A) Download all patient data locally and train models manually
B) Use BigQuery for structured data, Cloud Storage for unstructured data, and Vertex AI Pipelines for preprocessing, training, and deployment
C) Store data in spreadsheets and manually compute readmission risk
D) Train a model once using sample data and deploy permanently

Answer: B

Explanation:

Downloading all patient data locally and training models manually is unsuitable due to privacy, compliance, and scalability concerns. EHR data is sensitive and regulated by HIPAA and other regulations, and local storage increases the risk of unauthorized access. Manual training workflows are slow, error-prone, and non-reproducible. They cannot efficiently handle heterogeneous datasets that include structured EHR data, lab results, and imaging. Preprocessing manually is inconsistent and prevents automated retraining, which is essential to maintain predictive accuracy. This approach is inadequate for production healthcare predictive systems.

Using BigQuery for structured data, Cloud Storage for unstructured data, and Vertex AI Pipelines for preprocessing, training, and deployment is the most appropriate solution. BigQuery efficiently stores structured EHR data such as demographics, lab results, and medication histories, enabling large-scale querying and aggregation. Cloud Storage handles unstructured data such as clinical notes and imaging, allowing secure, scalable access. Vertex AI Pipelines orchestrate preprocessing, feature extraction, training, and deployment reproducibly, ensuring consistent handling of heterogeneous data. Continuous retraining pipelines allow models to adapt to new patient data, maintaining predictive accuracy. Logging, monitoring, and experiment tracking ensure operational reliability, reproducibility, and compliance with privacy regulations. Autoscaling supports processing large datasets efficiently. This architecture provides secure, scalable, reproducible, and continuously adaptive readmission risk predictions.

Storing data in spreadsheets and manually computing readmission risk is impractical. Spreadsheets cannot handle large-scale heterogeneous datasets, and manual computation is slow, error-prone, and non-reproducible. Continuous retraining and low-latency predictions are not feasible in this approach.

Training a model once using sample data and deploying permanently is inadequate. Patient populations, clinical procedures, and hospital conditions evolve over time, and a static model cannot adapt, reducing predictive accuracy. Continuous retraining and automated pipelines are essential for operational effectiveness.

The optimal solution is BigQuery for structured data, Cloud Storage for unstructured data, and Vertex AI Pipelines for reproducible, scalable, privacy-compliant readmission predictions.

Question 147

A retailer wants to forecast daily product demand across hundreds of stores using historical sales, promotions, holidays, and weather. The system must scale to millions of records, support feature reuse, and continuously update forecasts. Which solution is most appropriate?

Answer: B

Explanation:

Training separate models locally for each store using spreadsheets is impractical. Retail datasets involve millions of records across hundreds of stores and multiple feature types, including sales history, promotions, holidays, and weather. Local training cannot efficiently process this scale, is slow, error-prone, and non-reproducible. Managing features separately for each store introduces redundancy and inconsistency, and automated retraining pipelines are difficult to implement locally. Feature reuse is limited. This approach is unsuitable for enterprise-scale demand forecasting.

Using Vertex AI Feature Store for centralized features and Vertex AI Training for distributed forecasting is the most appropriate solution. Feature Store ensures consistent, reusable features across multiple models, reducing duplication and ensuring consistency between training and serving. Vertex AI Training supports distributed training across GPUs or TPUs, efficiently handling millions of historical records while capturing complex patterns across promotions, holidays, and weather. Pipelines automate feature updates, retraining, and versioning, ensuring forecasts continuously improve as new sales and promotion data becomes available. Autoscaling allows efficient handling of high data volumes. Logging, monitoring, and experiment tracking provide reproducibility, operational reliability, and governance compliance. This architecture enables scalable, accurate, and continuously updated demand forecasts across multiple products and stores.

The optimal solution is Vertex AI Feature Store combined with Vertex AI Training, providing scalable, reusable, and continuously updated inventory demand forecasts.

Question 148

A financial services company wants to implement real-time fraud detection for credit card transactions. The system must handle thousands of transactions per second, provide low-latency scoring, and adapt continuously to new fraud patterns. Which architecture is most appropriate?

A) Batch process transactions daily and manually review suspicious activity
B) Use Pub/Sub for transaction ingestion, Dataflow for feature engineering, and Vertex AI Prediction for online scoring
C) Store transactions in spreadsheets and manually compute fraud risk
D) Train a fraud detection model once per year and deploy it permanently

Answer: B

Explanation:

Batch processing transactions daily and manually reviewing suspicious activity is inadequate for real-time fraud detection. Fraud can occur within seconds, and daily batch processing introduces delays that allow fraudulent transactions to go undetected. Manual review cannot scale to handle thousands or millions of transactions per second and is prone to human error. Batch workflows also lack continuous retraining, preventing models from adapting to evolving fraud patterns. As a result, this approach is unsuitable for operational financial services requiring immediate fraud prevention.

Using Pub/Sub for transaction ingestion, Dataflow for feature engineering, and Vertex AI Prediction for online scoring is the most appropriate solution. Pub/Sub ingests credit card transactions continuously in real time, ensuring no transactions are missed. Dataflow pipelines compute derived features such as spending velocity, unusual transaction locations, device behavior patterns, and transaction frequency. Vertex AI Prediction provides low-latency scoring, enabling immediate fraud detection and alerts. Continuous retraining pipelines allow models to incorporate new fraud patterns and improve accuracy over time. Autoscaling ensures that high transaction volumes are handled efficiently. Logging, monitoring, and reproducibility provide operational reliability, auditability, and compliance with financial regulations. This architecture ensures scalable, low-latency, and continuously adaptive fraud detection.

Storing transactions in spreadsheets and manually computing fraud risk is impractical. Spreadsheets cannot handle high-frequency, high-volume transaction data. Manual computation is slow, error-prone, non-reproducible, and unsuitable for real-time operational fraud detection.

Training a fraud detection model once per year and deploying it permanently is insufficient. Fraud patterns evolve rapidly, and a static model cannot adapt, leading to decreased accuracy and potential financial loss. Continuous retraining and real-time scoring are essential for operational effectiveness.

Question 149

A logistics company wants to forecast delivery times in real time using vehicle telemetry, traffic data, and weather conditions. Predictions must scale to thousands of deliveries, provide low latency, and continuously adapt to changing conditions. Which solution is most appropriate?

Answer: B

Explanation:

Batch processing delivery data daily and manually updating predictions is inadequate for real-time logistics forecasting. Delivery times are influenced by dynamic factors such as traffic congestion, vehicle speed, and weather conditions that change continuously. Daily batch updates introduce delays, making predictions outdated and operationally ineffective. Manual updates cannot scale to thousands of vehicles and are prone to errors. Without continuous retraining and real-time feature computation, the predictive accuracy of the system degrades over time, limiting operational effectiveness.

Using Pub/Sub for real-time data ingestion, Dataflow for feature computation, and Vertex AI Prediction for online forecasting is the most appropriate solution. Pub/Sub ingests telemetry data, traffic updates, and weather conditions in real time. Dataflow pipelines compute derived features such as congestion impact, route delays, and estimated vehicle speed. Vertex AI Prediction provides low-latency forecasts to operational systems, enabling immediate adjustments to routes and delivery schedules. Continuous retraining pipelines allow models to adapt to evolving traffic patterns, vehicle behavior, and environmental conditions, improving forecast accuracy over time. Autoscaling supports peak delivery volumes efficiently. Logging, monitoring, and reproducibility provide operational reliability, traceability, and compliance. This architecture supports scalable, low-latency, continuously adaptive delivery time predictions.

Storing delivery data in spreadsheets and manually estimating delivery times is impractical. Spreadsheets cannot process high-volume, high-frequency telemetry and traffic data. Manual computation is slow, error-prone, non-reproducible, and unsuitable for real-time operational logistics.

Training a model once on historical data and deploying it permanently is insufficient. Delivery patterns change due to traffic, weather, and operational variations. Static models cannot adapt, leading to inaccurate predictions. Continuous retraining and real-time feature computation are necessary to maintain predictive accuracy.

Question 150

A retailer wants to forecast product demand across multiple stores using historical sales, promotions, holidays, and weather. The system must scale to millions of records, allow feature reuse, and continuously update forecasts. Which solution is most appropriate?

Answer: B

Explanation:

Training separate models locally for each store using spreadsheets is impractical. Retail datasets include millions of records across multiple stores and product types, covering features such as sales history, promotions, holidays, and weather. Local spreadsheets cannot handle this volume efficiently. Manual training is slow, error-prone, and non-reproducible. Managing features separately for each store introduces redundancy and inconsistency. Automated retraining pipelines are difficult to implement locally, and feature reuse is limited. This approach is unsuitable for enterprise-scale demand forecasting.

Using Vertex AI Feature Store for centralized features and Vertex AI Training for distributed forecasting is the most appropriate solution. Feature Store ensures consistent, reusable features across multiple models, reducing duplication and ensuring consistency between training and serving. Vertex AI Training supports distributed training across GPUs or TPUs, efficiently processing millions of historical records while capturing complex patterns in sales, promotions, holidays, and weather. Pipelines automate feature updates, retraining, and versioning, ensuring forecasts continuously improve as new data becomes available. Autoscaling allows efficient handling of high data volumes. Logging, monitoring, and experiment tracking provide reproducibility, operational reliability, and governance compliance. This architecture supports scalable, accurate, and continuously updated demand forecasts across multiple products and stores.

Storing historical sales data in Cloud SQL and training a single global linear regression model is insufficient. Cloud SQL is not optimized for large-scale analytical workloads, and linear regression cannot capture complex nonlinear relationships. A single model may underfit, producing inaccurate forecasts and lacking localized store-level precision.

Using a simple rule-based system based on last year’s sales is inadequate. Rule-based approaches cannot account for promotions, holidays, weather, or changing trends. They lack scalability, automation, and predictive accuracy, making them unsuitable for enterprise-level forecasting.

The optimal solution is Vertex AI Feature Store combined with Vertex AI Training, providing scalable, reusable, and continuously updated demand forecasts across multiple stores.

Google Professional Machine Learning Engineer Exam Dumps and Practice Test Questions Set 10 Q136-150

Related posts: