AI-900 Practice Question: Describe fundamental principles of machine learning on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Designing the hardware chips for running ML models

  Why it's wrong here

  Hardware design is microarchitecture and silicon engineering—it involves designing chips like GPUs, TPUs, or FPGAs to accelerate matrix multiplication and tensor operations. Feature engineering, by contrast, is a data-centric task performed before model training: it reshapes, encodes, scales, or derives new columns from raw data so the model can more easily detect predictive patterns. Choosing chip specifications has no direct effect on the information content of the training data, whereas feature engineering directly alters what the algorithm sees.

• ✓

  Selecting, transforming, and creating input variables from raw data to improve model performance

  Why this is correct

  Feature engineering is the deliberate process of selecting, transforming, and creating input variables (features) from raw data to make patterns more accessible to a machine learning algorithm. It includes techniques such as one-hot encoding categorical variables, normalizing numeric ranges, binning continuous values, extracting date/time components, and constructing interaction terms—all aimed at improving model accuracy, convergence speed, and generalization. This is the correct definition because it emphasizes the data representation transformation that precedes model training, distinct from hardware, deployment, or architecture choices.

• ✗

  Selecting which neural network layers to include in a model

  Why it's wrong here

  Selecting which neural network layers to include (e.g., convolutional, recurrent, dense, or attention layers) is neural architecture search or deep learning model design—it defines the network's topology, depth, and parameter capacity. Feature engineering, however, operates on the input data itself, independent of the model architecture: it decides what variables or transformations to feed into the network. While feature engineering can matter even for deep learning (e.g., creating time-based features or encoding cyclical patterns), the layer selection process is about the model's internal structure, not about reshaping raw inputs.

• ✗

  Writing code to deploy ML models as REST APIs

  Why it's wrong here

  Writing code to deploy ML models as REST APIs is model serving or MLOps engineering—it involves exposing a trained model's inference function through an HTTP endpoint, often using frameworks like FastAPI, Flask, or Azure Functions. This step happens after the model has been trained and validated, and it concerns infrastructure, request handling, serialization, and versioning. Feature engineering, in contrast, is the upstream, offline preparation of raw data into a clean, informative feature matrix that is used during both training and inference—so API deployment addresses how the model is exposed, not how the data is transformed.

Go deeper

Related to this question