Back to all pathsExplore GCP Guide Explore GCP Guide Explore GCP Guide
advanced
ML Engineer & Generative AI Path
Build, fine-tune, and deploy foundation models and custom neural networks.
35 mins3 Architectural Steps
STEP 1
Tabular ML & Feature Engineering in SQL
Train baseline predictive models directly in your data warehouse using standard SQL.
Google Cloud Platform
BigQuery ML (BQML)
Eliminates the need to export data into Python/R by executing linear regression, XGBoost, and k-means inside BigQuery.
Key Concepts:
CREATE MODEL statementAutoML Tables integrationVertex AI model export
CLI Example:
bq query --use_legacy_sql=false 'CREATE MODEL `proj.ds.churn_model` OPTIONS(model_type="logistic_reg") AS SELECT ...'STEP 2
Foundation Models & Generative AI Studio
Prompt engineer, ground, and fine-tune multimodal LLMs (Gemini, Claude, Llama).
Google Cloud Platform
Vertex AI Studio & Model Garden
Access Gemini 1.5 Pro/Flash with 1M+ token context windows, enterprise grounding with Google Search, and Vector Search.
Key Concepts:
Gemini 1.5 Multimodal APIModel Garden (Llama, Claude)Vector Search (Matching Engine)Enterprise Grounding
CLI Example:
gcloud ai models list --region=us-central1STEP 3
End-to-End MLOps & Continuous Training
Automate repeatable ML pipelines: data validation, training, model evaluation, and deployment.
Google Cloud Platform
Vertex AI Pipelines
Runs serverless Kubeflow Pipelines (KFP) and tracks dataset lineage, evaluation metrics, and model artifacts in Vertex ML Metadata.
Key Concepts:
Kubeflow Pipelines (KFP) SDKVertex ML MetadataServerless Pay-per-Second ExecutionLineage Graphs
CLI Example:
gcloud ai pipeline-runs list --region=us-central1