Professional Machine Learning Engineer Practice Questions

AutoML on Gemini Enterprise Agent Platform trains tabular classification models directly from BigQuery through the console: it shows data statistics, applies feature transformations, selects and tunes the model architecture automatically, and lets you deploy the model for serving, all without writing code. BigQuery ML requires writing SQL, and notebooks or custom training jobs require code.

The job runs on a single modest CPU machine. Giving the worker pool more powerful compute (a larger machine type or a GPU accelerator) shortens training time without changing the model or its hyperparameters, so accuracy is not affected. Reducing epochs or changing the batch size or learning rate alters the training dynamics and can significantly reduce accuracy.

Because riders confirm their presence and station one day in advance, passenger demand is already known and nothing needs to be predicted. The task is an optimization problem: compute the shortest route through all stations with confirmed attendance under capacity constraints and dispatch an appropriately sized shuttle. Rules of ML: do not use ML when a simple deterministic solution solves the problem; regression, classification or RL models would only approximate known data.

An experiment groups the runs of different model architectures, records their parameters, metrics and artifacts, and lets you compare the evaluation metrics of all runs side by side in one view (Experiments on Gemini Enterprise Agent Platform; Kubeflow Pipelines experiments serve the same purpose). AutoML does not build your custom Keras architectures, Composer only schedules jobs, and similarly named training jobs provide no shared comparison dashboard.

Horovod is a popular open-source distributed deep learning training framework that works with PyTorch, TensorFlow, and other frameworks. It is well-suited for data parallelism and is supported on Agent Platform. It simplifies the process of scaling a single-GPU program to run on multiple GPUs across multiple nodes.

Predicting inventory from region, location, historical demand and seasonal popularity is a supervised forecasting problem over time-ordered data. Recurrent neural networks (for example, LSTMs) model temporal dependencies such as seasonality and trends, and the model can be retrained daily on the newest inventory data. Reinforcement learning learns a policy for taking actions from reward signals rather than predicting a labeled quantity, classification predicts discrete classes rather than a demand quantity, and CNNs are designed for spatial (image-like) data.

Low loss on the training data with worse performance on the validation data means that the model is overfitting. L2 regularization and dropout are the standard ways to reduce overfitting, and their best values are found by experiment. So run a hyperparameter tuning job on Agent Platform (formerly Vertex AI) that searches over the L2 regularization and dropout parameters. Choosing one fixed dropout or L2 value and reducing the learning rate is guesswork, and the learning rate isn't a regularization setting. Tuning only the learning rate and doubling the number of neurons adds model capacity, which makes overfitting worse.

Ticket priority and resolution time are organization-specific predictions that require custom models trained on your own ticket history, so they are served from Gemini Enterprise Agent Platform (formerly Vertex AI) endpoints. Sentiment analysis of general-language text with no domain-specific jargon is exactly what the pre-trained Cloud Natural Language API provides, so no custom sentiment model needs to be trained or maintained. Vision models do not apply to text tickets.

Fraud is the rare (less common) class. For AutoML tabular classification, the AUC PR objective "maximize the area under the precision-recall curve" optimizes results for inferences for the less common class, balancing precision (few false positives) and recall (catching fraud). AUC ROC is the binary default and weighs both classes, log loss optimizes probability calibration, and precision at a fixed recall of 0.50 would miss half of the fraud.

Dataflow can process more than a million calls per day at scale in the originating region and can call Sensitive Data Protection (DLP) to de-identify PII before storage. BigQuery provides regional datasets and GoogleSQL, an ANSI-compliant SQL dialect, for the third-party visualization tool. Pub/Sub is messaging rather than processing, Datastore has no ANSI SQL interface, and Cloud SQL and Cloud Run functions are not designed for analytics at this scale.