Domain 2 of 6 · Chapter 5 of 5

Prebuilt AI APIs

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Professional Cloud Architect premium course — no separate purchase.

Included in this chapter:

  • The buy-build-tune ladder
  • Pre-trained APIs: one input, one result
  • Prebuilt solutions over your own data
  • Model Garden: one catalog, three model families
  • Exam-pattern recognition

Choosing a rung on the buy-build-tune ladder

Decision axisPre-trained APIPrebuilt solutionModel Garden modelCustom model on Vertex AI
ExamplesCloud Vision, Speech-to-Text, Cloud Translation, Natural Language, Document AIAI Applications (Agent Search), Conversational Agents, Gemini EnterpriseGemini, Imagen, Claude, Gemma, LlamaAutoML or custom-trained model
What you supplyJust the input (image, audio, text)Your data store plus light configA prompt, optionally tuning dataLabelled training data and pipelines
ML expertise neededNoneLowLow to mediumHigh (data scientists)
DifferentiationNone (shared model)Your data, shared modelSome (prompt or tuning)Highest (your model)
Typical taskFixed perception or transcriptionSearch and chat over your corpusGenerative or reasoningUnique prediction on your data

Decision tree

Generic pre-trained APIcovers the task?Pre-trained APIVision, Speech, TranslationyesGrounded in yourown corpus?noPrebuilt solutionAI Applications, Conv. AgentsyesGenerative or reasoning,solved by a prompt?noModel Garden modelGemini API or open modelyesCustom model on Vertex AImodel learns your patternsno
Cheat sheet

Cheat sheet

  • Climb the buy-build-tune ladder only as far as differentiation needs
  • A pre-trained API is a stateless model call: you select and call, never train
  • Use the Cloud Vision API to read images, Document AI to structure documents
  • Cloud Translation Advanced adds Adaptive Translation, the lightest tuning on this page
  • Speech-to-Text V2 lets you choose a model like Chirp without training one
  • For grounded answers over your own corpus, use AI Applications (Agent Search)
  • AI Applications has many former names; treat them all as one product
  • For a multi-turn dialogue, use Conversational Agents (Dialogflow CX), not a search box
  • Gemini Enterprise packages AI agents and NotebookLM for staff, not developers
  • NotebookLM grounds its answers only on the sources you give it
  • Reach every foundation model, including Gemini, through Vertex AI Model Garden
  • Managed-API models need no serving infra; open models are self-deployed to an endpoint
  • When data cannot leave your boundary, self-deploy an open model in your project
  • Gemini powers the prebuilt solutions; pick the layer by how much you must build
  • If the model must learn your patterns, leave the prebuilt rung for custom training

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. Cloud Vision API documentation
  2. AutoML
  3. Google Cloud Digital Leader certification exam guide
  4. Document AI overview
  5. Cloud Speech-to-Text documentation
  6. Text-to-Speech documentation
  7. Cloud Translation overview
  8. Cloud Natural Language API documentation
  9. AI Applications introduction
  10. Conversational Agents (Dialogflow CX) agent concept
  11. Gemini Cloud Assist overview
  12. Explore models in Vertex AI Model Garden