Domain 3 of 6 · Chapter 3 of 3

Building AI/ML Solutions

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

Bundled into the existing Cloud Digital Leader premium course — no separate purchase.

Included in this chapter:

  • The build ladder: four ways to get ML into a product
  • BigQuery ML: training models in SQL
  • Pre-trained APIs: ready-made models by data type
  • AutoML, Vertex AI, TensorFlow and Cloud TPUs
  • Exam-pattern recognition

Pre-trained ML APIs by data type and use case

APIData typeWhat it does / when to use
Cloud Vision APIImagesUnderstands image content: labels, OCR/text, faces, objects, logos, landmarks, explicit-content detection
Cloud Natural Language APITextUnderstands text: sentiment, entities, syntax, and content classification
Cloud TranslationTextTranslates text between languages and detects the source language
Speech-to-TextAudioTranscribes spoken audio into written text (e.g. call transcription, voice input)
Text-to-SpeechTextSynthesises natural-sounding speech audio from text (e.g. voice responses, narration)

Decision tree

Does a pre-trained APIsolve this generic task?YesPre-trained APIVision / NL / Translation / SpeechNoAnalysts with data inBigQuery, using SQL?YesBigQuery MLNoHave your own labelled databut no ML team?YesAutoMLNoCustom model on Vertex AIyour training code (e.g. TensorFlow)Always: pick the lowest-effort rung that solves the problemeffort and differentiation rise as you climb
Cheat sheet

Cheat sheet

  • Climb the build ladder only as high as the problem needs
  • BigQuery ML lets analysts train models in SQL, where the data already lives
  • BigQuery ML covers the everyday model types in SQL
  • Pre-trained APIs need no training data and no ML skill
  • Match the pre-trained API to the data type
  • The Vision API reads text in images with OCR. No custom model needed
  • The Natural Language API gauges sentiment and entities in text
  • Speech-to-Text and Text-to-Speech are inverses. Watch the direction
  • Cloud Translation converts text between languages with no training
  • AutoML trains a custom model on your data without you building the model
  • AutoML still needs your own labelled data
  • Vertex AI is Google Cloud's single unified ML platform
  • Build a custom model only when the model is your competitive edge
  • TensorFlow is open-source software; a Cloud TPU is Google's proprietary hardware
  • A Cloud TPU is a custom ASIC built to accelerate ML
  • Dialogflow is Google Cloud's product for building virtual agents and chatbots
  • Vertex AI manages models after training: evaluate, deploy/version, and monitor for drift

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

Also tested in

References

  1. https://docs.cloud.google.com/bigquery/docs/bqml-introduction
  2. https://docs.cloud.google.com/vertex-ai/docs/start/introduction-unified-platform
  3. https://docs.cloud.google.com/vision/docs/features-list
  4. https://docs.cloud.google.com/natural-language/docs/basics
  5. https://docs.cloud.google.com/translate/docs/overview
  6. https://docs.cloud.google.com/speech-to-text/docs/basics
  7. https://docs.cloud.google.com/text-to-speech/docs/basics
  8. https://docs.cloud.google.com/tpu/docs/intro-to-tpu