Cloud-Based Machine Learning Services (SageMaker, AI Platform) Questions and Answers for Viva

Answer-1: AWS SageMaker is a fully managed service that provides developers and data scientists the ability to build, train, and deploy machine learning models at scale.

Answer-2: The main components include SageMaker Studio, Notebooks, Training, Tuning, Models, and Endpoints for deployment.

Answer-3: It automates infrastructure management and provides one-click deployment to hosted endpoints for real-time inference.

Answer-4: Google Cloud AI Platform is a managed service that enables building, deploying, and managing machine learning models on Google Cloud.

Answer-5: AI Platform supports distributed training on scalable infrastructure using custom or pre-built containers.

Answer-6: SageMaker supports TensorFlow, PyTorch, MXNet, Chainer, and scikit-learn, among others.

Answer-7: Ground Truth is a data labeling service that uses machine learning and human annotators to create high-quality labeled datasets.

Answer-8: It is an automated process to find the best hyperparameters for a model to improve accuracy and performance.

Answer-9: It provides an integrated development environment with tools for data preparation, model training, debugging, and deployment.

Answer-10: It enables automated ML workflows for continuous integration and delivery of ML models.

Answer-11: SageMaker uses AWS IAM roles, VPCs, encryption at rest and in transit, and integrates with AWS KMS for key management.

Answer-12: Batch transform processes large datasets asynchronously, while real-time endpoints provide immediate predictions for live data.

Answer-13: It supports model versioning by storing multiple model versions and managing deployment versions.

Answer-14: SageMaker Neo optimizes machine learning models to run faster and with lower latency on multiple hardware platforms.

Answer-15: It automatically builds, trains, and tunes machine learning models from raw data without requiring deep ML expertise.

Answer-16: Pricing is based on usage of notebook instances, training jobs, model hosting, and data processing.

Answer-17: AI Platform allows users to bring their own Docker containers for customized training and prediction environments.

Answer-18: Endpoints are deployed instances of trained models that serve predictions via APIs.

Answer-19: SageMaker automatically scales infrastructure for training and hosting based on workload demand.

Answer-20: Yes, it integrates with BigQuery, Cloud Storage, AI Hub, and Dataflow among others.

Answer-21: Debugger monitors training jobs in real-time to detect and alert on issues like overfitting or vanishing gradients.

Answer-22: Using CloudWatch metrics, SageMaker Model Monitor, and logging to track performance and data drift.

Answer-23: TFX provides a platform for deploying production ML pipelines integrated with AI Platform.

Answer-24: Yes, SageMaker supports reinforcement learning frameworks and provides RL-specific algorithms.

Answer-25: It allows preprocessing using custom code in training jobs or via AI Platform Pipelines.

Answer-26: A service to run data processing and model evaluation workloads in managed containers.

Answer-27: Through IAM roles, encryption, VPC Service Controls, and private network connectivity.

Answer-28: JumpStart offers pre-built solutions, example notebooks, and models to accelerate ML projects.

Answer-29: By uploading a trained model to Cloud Storage and creating a model resource and version on AI Platform.

Answer-30: Managed Jupyter notebooks integrated with Google Cloud for easy data science and ML development.

Answer-31: It allows hosting multiple models on a single endpoint to reduce cost and improve resource utilization.

Answer-32: Clarify provides tools for detecting bias and explaining model predictions to improve transparency.

Answer-33: By configuring distributed training jobs with multiple worker nodes and parameter servers.

Answer-34: It tracks, compares, and organizes machine learning training runs and parameters.

Answer-35: It integrates with tools like Explainable AI to interpret model predictions.

Answer-36: Python is the primary language supported for SageMaker SDK and APIs.

Answer-37: For data cataloging and ETL workflows to prepare data for machine learning.

Answer-38: TensorFlow, scikit-learn, XGBoost, PyTorch, and custom models packaged in containers.

Answer-39: By submitting batch prediction jobs that process large datasets stored in Cloud Storage.

Answer-40: It helps manage model versions, approvals, and deployment stages.

Answer-41: Using SageMaker Pipelines to create end-to-end ML workflows.

Answer-42: Training is for model development, while Prediction is for serving the trained model.

Answer-43: By choosing spot instances for training, using multi-model endpoints, and deleting unused resources.

Answer-44: Pre-optimized machine learning algorithms provided by AWS for common use cases like classification and regression.

Answer-45: Via continuous evaluation pipelines and integration with monitoring tools.

Answer-46: Yes, it supports deep learning frameworks and GPU instances for training.

Answer-47: Using Cloud Storage versioning and dataset management tools integrated with pipelines.

Answer-48: Data preparation ? Model building ? Training ? Tuning ? Deployment ? Monitoring.

Answer-49: Through Vertex AI AutoML services that automate model training and tuning.

Answer-50: Use encryption, IAM policies, VPCs, audit logs, and secure data access controls.