Free Practice Questions for Snowflake MLA-B01 Certification

● Use ML preprocessing functions, (e.g., MinMaxScaOneHotEncoder) to transform large data sets within Snowflake ● Load and transform data with DataFrames ● Analyze use cases to determine when to use SQL vs Snowpark ● Select appropriate data storage methods for training and inference datasets (e.g., external/internal stages, tables, data shares) based on performance, cost, and governance requirements

● Define and configure Feature Store entities to establish data models for ML features ● Create feature views to encapsulate complex transformations and provide consistent interfaces ● Implement Feature Store architecture to enable centralized feature management and reuse ● Register ML features as centralized metadata objects to enable discovery and reuse across multiple use cases and teams ● Configure external feature views to integrate and standardize features from data sources outside Snowflake while maintaining Feature Store consistency ● Track versions of features using version control ● Use Snowpark dataframe to build out a feature store ● Identify how dynamic tables are used in the Feature Store ● Manage data sets for training, validation, and inference across ML workflows

● Design lookups using Snowflake Feature Store views to generate training sets that prevent temporal data leakage ● Verify consistency between offline training sets and low-latency online feature retrieval ● Ensure Dev/Prod feature consistency

● Snowflake Lineage visualization and querying ● Configure streaming data ingestion for near real-time feature generation ● Implement Snowflake Data Quality Monitoring to validate data integrity at ingestion ● Configure feature validation rules beyond basic data quality checks ● Monitor feature freshness, data quality, schema changes, and drift using Snowflake observability tool

● Package and version feature transformations using Snowpark, SQL, or dynamic tables independently from model code ● Scale batch and near-real-time feature computation using Snowflake warehouses or compute pools ● Deploy and promote feature pipelines across Dev/Test/Prod using tasks and dynamic table refresh policies with offline/online parity

● Deploy custom model runtime environments using containerized images to support libraries not natively available in Snowpark ● Operate ML Jobs ● Configure and compare ML distributed APIs across warehouse vs container runtime environments ● Configure compute pools for ML workloads and Snowflake Container Services ● Optimize resource allocation between warehouses and Container Runtime

● Train models within Snowflake Notebooks in Workspaces to leverage distributed data processing capabilities ● Analyze use cases to determine when to develop in Workspaces vs stored procedures, SQL, or Python files ● Use open-source packages to build and evaluate models ● Use DataConnector with open source packages to speed up data ingestion for ML work ● Identify the default GPU allocation constraints for individual notebook instances within the Workspaces environment

● Structure nested experiments to log performance metrics, hyperparameters, and artifacts using Snowpark ML metadata ● Execute hyperparameter optimization (HPO) workflows within Snowflake's ML framework ● Implement CustomModel API for integrating external models and custom algorithms into the Snowflake ML framework ● Monitor comprehensive experiment tracking for model development lifecycle management ● Associate models with specific dataset versions to ensure full lineage and experiment reproducibility

● Register model artifacts, custom models, versions, and aliases with proper metadata and dependencies ● Archive and deprecate model versions ● Execute model rollback procedures when performance degrades ● Assign model aliases (e.g., @prod) to decouple client applications from specific underlying model versions ● Extract and manage model artifacts including pickle files and other serialized model components ● Validate model artifact integrity during registration and deployment processes ● Deploy registered models across different environments and accounts ● Configure model promotion workflows between dev/test/prod environments ● Validate model compatibility across environment configurations

● Deploy models to managed HTTPS endpoints in Snowpark Container Services for real-time, low-latency inference ● Execute batch inference pipelines using registry-backed execution to process large datasets within dedicated virtual warehouses ● Deploy batch inference as service functions using the same service used for online model deployment ● Run large batch workloads on Snowpark Container Services using job-based batch inference (without service creation) ● Process unstructured data and multimodal models through batch inference capabilities ● Track metrics for Real-time Inference via REST API endpoints

● Migrate existing models from third-party platforms into Snowflake ● Identify Snowflake-native alternatives to external MLOps tools and workflows

● Develop automated pipelines that chain data validation, training, and deployment ● Integrate ML capabilities with developer tools including Snowflake CLI, SnowSQL, and language-specific SDKs

● Configure Git integration in Snowflake to automate ML code deployment ● Leverage Snowflake CLI for automation ● Deploy ML assets using Snowflake-native promotion workflows ● Configure Snowflake object dependencies for ML pipeline deployment

● Define automated retraining policies triggered by data drift alerts or performance degradation metrics ● Troubleshoot pipeline failures related to warehouse capacity, container scalability, or data dependency issues

● Ensure appropriate privileges are granted to roles for accessing Snowflake ML components (e.g., Feature Store, Model Registry, or compute pools) ● Apply Snowflake Dynamic Data Masking and row-level Security to ensure sensitive data is protected during training and inference ● Establish data governance policies for ML model compliance across feature stores and model registries

● Track data drift and statistical anomalies using Snowflake's integrated monitoring and alerting ● Audit all model interactions and inference calls to maintain lineage for regulated industries ● Configure alerts for ML data governance violations ● Configure automated model performance degradation alerts ● Implement model accuracy and drift threshold monitoring ● Monitor model serving performance metrics (latency, throughput, availability) to enhance operational monitoring coverage

● Tune warehouse optimization strategies and compute pool sizing to balance ML performance with cost efficiency ● Track costs for ML features with granular attribution ● Track costs of Snowpark Container Services compute pool and optimize it if necessary