Architecting low-code AI solutions — PMLE Practice Quiz

Question 1

A retail chain wants to classify products on store shelves using cameras that operate in stores with unreliable internet connectivity. The data science team has limited ML coding experience but has thousands of labeled product images. They need a solution that trains a custom image classifier and can run inference directly on in-store edge devices without a constant network connection. Which approach best meets these requirements?

Accepted Answer

Train a custom model with Vertex AI AutoML Image Classification using the 'Edge' model type, then export the model to run on the in-store devices.

Answer

Use the Cloud Vision API label detection endpoint from each in-store device to classify products in real time.

Answer

Train a Vertex AI AutoML Image Classification model with the 'Cloud' hosted model type and call its online prediction endpoint from the devices.

Answer

Build a custom TensorFlow CNN in Vertex AI Workbench notebooks and containerize it for edge deployment.

Question 2

A fintech company wants to build a binary classification model in Vertex AI AutoML Tabular to detect fraudulent transactions. Only 0.8% of the labeled transactions are fraudulent. The team's priority is to catch as many actual fraud cases as possible while keeping false positives at a manageable level, and they have limited ML expertise. Which approach should they configure to best meet this goal?

Accepted Answer

Train the AutoML Tabular model and set the optimization objective to maximize the area under the precision-recall curve (AU PRC).

Answer

Train the AutoML Tabular model with the default optimization objective of maximizing the area under the ROC curve (AU ROC).

Answer

Manually downsample the majority class to a 50/50 balance before training and optimize for overall accuracy.

Answer

Switch to a BigQuery ML logistic regression model and optimize for log loss since AutoML cannot handle imbalanced data.

Question 3

A logistics company wants to predict package delivery times using a structured dataset in Vertex AI. The data science team has limited coding experience and needs a solution that requires minimal ML expertise. Business stakeholders also require the model to report which features most influence each individual prediction, and the model must be queried in real time from a customer-facing web application. Which approach best meets all requirements?

Accepted Answer

Train an AutoML tabular regression model in Vertex AI, deploy it to an online endpoint, and enable Vertex Explainable AI feature attributions on the predictions.

Answer

Train a custom TensorFlow regression model with a hand-written training loop and serve it via a Compute Engine VM.

Answer

Train an AutoML tabular regression model in Vertex AI and use only batch prediction jobs to generate delivery time estimates.

Answer

Use BigQuery ML to train a linear regression model and export it to a TensorFlow SavedModel for local inference.

Question 4

A marketing analytics team with limited ML expertise wants to build a churn prediction model from a customer table already in BigQuery. They choose Vertex AI AutoML Tabular through the console. During job configuration, they must set a training budget expressed in node hours. They ask you how this setting affects the outcome and how to choose a reasonable value for their first experiment. What is the most accurate guidance?

Accepted Answer

The budget caps the maximum compute AutoML spends searching architectures and tuning; AutoML may stop early if the model converges, and a small budget is appropriate for an initial baseline experiment.

Answer

The budget sets the number of prediction nodes for the deployed endpoint, so it should match expected serving traffic rather than training needs.

Answer

The budget determines how many rows of the dataset AutoML uses for training, so a higher budget always means more data is included.

Answer

The budget is a hard minimum runtime; AutoML will always consume the entire allocation even if the model has already converged, so it should be set as low as possible to save money.

Question 5

A sports media company wants to automatically tag short game clips with the specific athletic actions they contain (e.g., 'shot on goal', 'free kick', 'tackle') along with the time segments where each action occurs. The data science team has only a few weeks, limited ML expertise, and a labeled dataset of a few thousand video clips. They want the lowest-code approach on Google Cloud that produces temporal action segments. Which solution best fits these requirements?

Accepted Answer

Use AutoML (Vertex AI) to train an action recognition video model on the labeled clips

Answer

Use the Vision API's label detection endpoint on sampled frames from each clip

Answer

Write a custom PyTorch 3D-CNN training job on Vertex AI custom training with a GPU cluster

Answer

Use the Video Intelligence API's prebuilt label detection to identify actions

Question 6

A retail analytics team stores all their customer transaction data in BigQuery. They want to build a churn prediction model directly in BigQuery ML using a boosted tree classifier, and they want BigQuery ML to automatically search for the best combination of learning rate and max tree depth rather than manually experimenting. They also need the final model to be usable for online predictions from a Vertex AI endpoint later. Which approach best meets these requirements with the least custom code?

Accepted Answer

Create a BOOSTED_TREE_CLASSIFIER model in BigQuery ML with NUM_TRIALS and HPARAM_RANGE/HPARAM_CANDIDATES options to enable hyperparameter tuning, then export the model and register it to Vertex AI for online serving.

Answer

Train a custom XGBoost model in a Vertex AI custom training job, manually tune hyperparameters with Vertex AI Vizier, then deploy to an endpoint.

Answer

Use AutoML Tables in Vertex AI, which handles hyperparameter tuning automatically, and skip BigQuery ML entirely.

Answer

Create a LOGISTIC_REG model in BigQuery ML with L1_REG and L2_REG regularization options, since boosted trees do not support hyperparameter tuning in BigQuery ML.

Question 7

A retail analytics team stores three years of daily sales transactions in BigQuery. They need to produce a 30-day demand forecast for each of 5,000 products, including confidence intervals, while keeping the data in place and minimizing custom code or infrastructure management. Which approach best meets these requirements?

Accepted Answer

Create a BigQuery ML ARIMA_PLUS model with a time-series column, item identifier, and TIME_SERIES_ID_COL to train per-product models, then use ML.FORECAST to generate predictions with prediction intervals.

Answer

Export the sales data to Cloud Storage, train a custom LSTM model on Vertex AI with a TensorFlow container, and deploy it to an endpoint for batch forecasting.

Answer

Use a BigQuery ML LINEAR_REG model with the date as a numeric feature and call ML.PREDICT to estimate future sales for each product.

Answer

Send each product's historical sales to the Vertex AI Gemini API and prompt the model to predict the next 30 days of demand.

Question 8

A retail analytics team has a BigQuery table with millions of rows of customer transaction data, including recency, frequency, and monetary value features. The marketing department wants to group customers into an unspecified number of natural segments for targeted campaigns, but there are no pre-existing labels indicating which segment each customer belongs to. The team wants a fully SQL-based, low-code approach without exporting data. Which BigQuery ML approach should they use?

Accepted Answer

Create a k-means model using CREATE MODEL with MODEL_TYPE='KMEANS' and evaluate cluster quality with the Davies-Bouldin index

Answer

Create a logistic regression model using MODEL_TYPE='LOGISTIC_REG' to predict each customer's segment label

Answer

Create a multiclass classification model using MODEL_TYPE='DNN_CLASSIFIER' on the transaction features

Answer

Create an ARIMA_PLUS time-series model to forecast future customer purchasing behavior

Question 9

A retail analytics team has trained a logistic regression model in BigQuery ML to predict which customers will churn. Business stakeholders want to understand, for each individual prediction, which features pushed the customer toward or away from churning so they can design targeted retention offers. The team wants to stay entirely within BigQuery using SQL and avoid exporting the model. Which approach should they use?

Accepted Answer

Run ML.EXPLAIN_PREDICT on the model to obtain per-prediction feature attributions alongside the predicted labels

Answer

Run ML.GLOBAL_EXPLAIN after enabling global explanations during training to see per-customer feature contributions

Answer

Run ML.EVALUATE to produce a confusion matrix and precision/recall values for each customer

Answer

Export the model to Vertex AI and enable Vertex Explainable AI with sampled Shapley to get per-row attributions

Question 10

A SaaS company stores millions of historical customer support tickets in BigQuery, each labeled with one of five routing categories (Billing, Technical, Account, Sales, Other). The data engineering team wants to automatically predict the routing category for new tickets using the text description column, and they prefer to build and serve the model without moving data out of BigQuery or writing custom training code. Which approach best meets these requirements?

Accepted Answer

Use BigQuery ML to train a model with MODEL_TYPE='LOGISTIC_REG' on the label column, letting BigQuery ML automatically detect the multiclass target and generate text feature transformations.

Answer

Create a BigQuery ML logistic regression model with MODEL_TYPE='LOGISTIC_REG', which natively handles the five-category target and text features.

Answer

Export the tickets to Cloud Storage and train a custom PyTorch text classifier on Vertex AI, then deploy it to an online endpoint.

Answer

Call the Cloud Natural Language API's classifyText method on each ticket to obtain the routing category directly.

Question 11

A retail analytics team stores millions of transaction records in BigQuery. A data scientist has already trained and deployed a custom fraud-scoring model to a Vertex AI online endpoint. The business analysts, who only know SQL, need to run predictions against this deployed model directly from their BigQuery queries without exporting data or writing Python. Which approach best meets this requirement with the least operational overhead?

Accepted Answer

Create a BigQuery ML remote model referencing the Vertex AI endpoint, then call ML.PREDICT on it from SQL.

Answer

Export the BigQuery table to Cloud Storage and run a Vertex AI batch prediction job against the endpoint.

Answer

Re-train the fraud model entirely inside BigQuery ML using CREATE MODEL so the analysts can use ML.PREDICT.

Answer

Have the analysts call the Vertex AI endpoint through a Cloud Function that they invoke via HTTP for each row.

Question 12

A retail analytics team stores millions of customer feedback comments in a BigQuery table. Analysts who know SQL but have no ML or Python experience want to classify each comment into one of five predefined product-category themes and generate a one-sentence summary, running the process entirely within their existing BigQuery workflows. They want the fastest path to production without training a custom model or standing up separate inference infrastructure. What should they do?

Accepted Answer

Create a BigQuery ML remote model that references a Vertex AI Gemini endpoint, then call ML.GENERATE_TEXT with a prompt column that instructs the model to classify and summarize each comment.

Answer

Export the comments to Cloud Storage, train an AutoML text classification model in Vertex AI, deploy it to an online endpoint, and call it from a scheduled query.

Answer

Use CREATE MODEL with model_type='logistic_reg' in BigQuery ML to train a multiclass classifier directly on the raw comment text.

Answer

Build a Vertex AI Pipeline that reads from BigQuery, fine-tunes a foundation model on the comments, and writes predictions back to a results table.

Question 13

A retail analytics team stores millions of product reviews in a BigQuery table. Analysts who are comfortable with SQL but not with Python want to generate concise one-sentence summaries of each review directly inside their existing BigQuery workflow, without building a custom pipeline or managing serving infrastructure. Which approach best meets these requirements with the least operational overhead?

Accepted Answer

Create a BigQuery ML remote model that references a Vertex AI Gemini endpoint, then call ML.GENERATE_TEXT in a SQL query over the reviews table.

Answer

Export the reviews to Cloud Storage, run a Vertex AI batch prediction job against a Gemini model, and load the summaries back into BigQuery.

Answer

Train a custom text summarization model with AutoML and deploy it to a Vertex AI online endpoint that analysts query via the REST API.

Answer

Build a Dataflow streaming pipeline that reads from BigQuery, calls the Vertex AI SDK for each row, and writes summaries to a new table.

Question 14

A data analytics team wants to run entity sentiment analysis on millions of product reviews stored in a BigQuery table. They prefer to stay entirely within SQL and avoid writing any Python or exporting data. They plan to use the Cloud Natural Language API via BigQuery ML. When creating the remote model with CREATE MODEL ... REMOTE WITH CONNECTION, what is the essential configuration step required so that BigQuery can successfully invoke the Cloud Natural Language API?

Accepted Answer

Grant the BigQuery connection's service account the appropriate IAM role (e.g., Cloud Natural Language API user/service usage) and enable the Cloud Natural Language API in the project

Answer

Export the reviews to Cloud Storage first, because BigQuery ML remote models can only read input from GCS buckets

Answer

Deploy the Natural Language model to a Vertex AI endpoint and reference that endpoint URL in the ENDPOINT option of the remote model

Answer

Create a service account key file and embed the JSON credentials directly in the SQL CREATE MODEL statement

Question 15

A data analytics team stores millions of customer review texts in a BigQuery table. They want to build a semantic search feature that finds reviews similar to a given query, and they prefer to stay entirely within SQL workflows without exporting data or writing Python pipelines. They need to generate text embeddings at scale directly in BigQuery. What is the most appropriate low-code approach?

Accepted Answer

Create a BigQuery ML remote model that references a Vertex AI text embedding foundation model, then call ML.GENERATE_EMBEDDING on the reviews table.

Answer

Train an AutoML text classification model in Vertex AI and use its prediction scores as embedding vectors for similarity search.

Answer

Export the reviews to Cloud Storage and call the Natural Language API's entity analysis to compute similarity between documents.

Answer

Use CREATE MODEL with a BigQuery ML matrix factorization model to produce embeddings from the review text columns.

Question 16

A retail analytics team stores 4 million product photos in Cloud Storage, with their metadata (GCS URIs, SKU, category) already loaded into a BigQuery object table. Analysts who are fluent in SQL but have no ML training pipeline experience need to generate descriptive labels for each image to enrich the product catalog. They want the lowest-effort approach that keeps everything within their existing BigQuery workflow and does not require them to build or manage any serving infrastructure. What should they do?

Accepted Answer

Create a BigQuery ML remote model that connects to the Cloud Vision API and call ML.ANNOTATE_IMAGE over the object table to generate labels directly in SQL.

Answer

Train an AutoML image classification model in Vertex AI, deploy it to an online endpoint, and have analysts write SQL that calls the endpoint per row.

Answer

Export all images to a Vertex AI managed dataset, run a custom TensorFlow training job to build a labeler, then batch predict and load results back to BigQuery.

Answer

Write a Dataflow pipeline that reads the object table, calls the Vision API for each image, and streams annotations into a BigQuery results table.

Question 17

A retail analytics team has an existing TensorFlow SavedModel that predicts product return likelihood. The model was trained outside Google Cloud by a data science partner. The team wants their SQL-focused analysts to run batch predictions on customer transaction data that already lives in BigQuery, without exporting the data or standing up a separate serving infrastructure. What is the most appropriate low-code approach?

Accepted Answer

Use CREATE MODEL with the MODEL_TYPE='TENSORFLOW' option and MODEL_PATH pointing to the SavedModel in Cloud Storage, then run ML.PREDICT against the BigQuery table

Answer

Deploy the SavedModel to a Vertex AI endpoint and have analysts call the endpoint from a Cloud Function for each transaction row

Answer

Retrain the model from scratch as a BigQuery ML boosted tree classifier since imported TensorFlow models cannot be used for prediction in BigQuery

Answer

Export the BigQuery table to Cloud Storage as CSV and run the TensorFlow model locally with a Python batch script

Question 18

A healthcare analytics company needs to extract structured data (patient ID, test names, numeric result values, and reference ranges) from thousands of scanned lab reports that come from dozens of different labs, each using its own unique layout. The pre-trained Document AI processors do not recognize these domain-specific fields, and the layouts vary too widely for a template-based rules approach. The team wants the lowest-code path to a production extraction pipeline. What should they do?

Accepted Answer

Use the Document AI Custom Extractor processor, label a set of representative lab reports with the target fields, and train the processor to extract them.

Answer

Use the Document AI Form Parser processor, which automatically extracts all key-value pairs from any document layout without training.

Answer

Use the Document OCR processor to extract raw text, then write custom regular expressions to locate each field in the output.

Answer

Fine-tune a Gemini model in Model Garden on the scanned images to output the structured fields as JSON.

Question 19

A healthcare clinic receives thousands of scanned patient intake forms each day. The forms use a consistent template with labeled fields (name, date of birth, insurance ID) but often contain handwritten patient entries. The team wants a low-code solution that extracts the field-value pairs into structured key-value data without building or training a custom model, and needs it to handle the handwriting reliably. Which Google Cloud approach should they choose?

Accepted Answer

Use the Document AI Form Parser processor, which extracts key-value pairs and includes handwriting recognition out of the box.

Answer

Use the Vision API DOCUMENT_TEXT_DETECTION feature and write custom logic to associate text blocks into field-value pairs.

Answer

Train a Document AI Custom Extractor processor with hundreds of labeled documents to recognize each field.

Answer

Use the Vision API TEXT_DETECTION feature to return dense text, then apply a regex pipeline to reconstruct the form fields.

Question 20

A logistics company receives thousands of supplier invoices daily as scanned PDFs with varying layouts. The finance team wants to automatically extract structured fields such as invoice number, vendor name, line items, and total amount to feed into their accounts payable system, with minimal custom ML development. Which Google Cloud approach best meets this requirement?

Accepted Answer

Use the Document AI Invoice Parser processor to extract the structured entities from the PDFs

Answer

Use the Cloud Vision API's text detection (OCR) feature and write regular expressions to parse the extracted text

Answer

Train a custom AutoML Vision image classification model to categorize the invoice images

Answer

Load the raw PDF bytes into BigQuery and use BigQuery ML to run a logistic regression on the documents

More PMLE practice