Compare Google Cloud’s business and engineering certifications and build a project with a measurable evaluation plan.

Generative AI Leader

Start here if you choose use cases, coordinate AI adoption, or sponsor delivery. Write a short business case with a baseline, a measurable outcome, and a plan for responsible use.
Professional Machine Learning Engineer

Choose this route if you build and operate AI or ML workloads. Practice data preparation, evaluation, deployment, monitoring, and investigation of performance changes.
Turn a business idea into a measurable pilot
Pick a narrow job such as answering onboarding questions. Record how long people spend finding answers today. Compare the pilot on citation accuracy, adoption, and time to a correct answer. Include a manual fallback and agree on what result would justify a wider rollout. This is a practical exercise for Generative AI Leader candidates.
Follow the model through its lifecycle
For Professional Machine Learning Engineer, create versioned training and evaluation sets. Use a reproducible training pipeline and register the model before deployment. Test a workload identity with deliberately denied permissions. Compare performance across data slices, decide what should trigger retraining, and practice restoring the previous model.
Use the same cases as the application changes
Create a fixed suite with correct answers, ambiguous questions, missing evidence, and instructions embedded in a source document. Compare Gemini responses with retrieval enabled and disabled. Keep the evaluation results with the selected model and prompt version so the next person can repeat the comparison.
Build something you can show
Build a Gemini assistant that answers questions about public documentation. Use the Gemini Enterprise Agent Platform, an evolution of Vertex AI, to explore model and agent capabilities. Give the application a narrowly scoped service identity. Compare retrieval and no-retrieval runs on a fixed question set, then record citation accuracy, latency, and cases where the assistant should abstain.