AI/ML Services

GCP AI/ML Services

AI Building Blocks

  • AI through simple REST calls

  1. Sight

    1. Vision API : Derive insights from images
    2. Video API : content discovery and engaging video experiences

  2. Conversation

    1. Dialogflow : to build virtual chat agents and conversational experiences
    2. Cloud Text-to-Speech API : to convert text to human like speech using Wavenet voices
    3. Cloud Speech-to-Text API : to convert speech to text

  3. Language

    1. Translation: to detect and translate between languages
    2. Natural Language: Reveal the structure and meaning of text through machine learning

  4. Structured Data

    1. AutoML Tables: build and deploy machine learning models on structured data
    2. Recommendations AI: delivers personalized recommendations
    3. Cloud Inference API: to run large-scale correlations over time-series datasets

Cloud AutoML

  • training models on custom data sets
  • custom machine learning models without writing code
  • based google’s ml algorithms
  • AutoML services
    • Sight
      • Vision
      • Video
    • Language
      • Natual Language
      • Translation
    • Struture Data
      • Tabular Data

AI Platform

  • end-to-end ML pipelines on-premises and cloud
  • targetted for users who are building custom ml models and want utilize ML pipelines
  • based on Kubeflow, open source ML project based on kubernetes

General steps in a ML pipeline

Ingest Data
	|
 Prepare
 	|
Pre-Process
	| 
 Discover
 	|
 Develop
 	|
  Train
  	|
Test & Analyze 
	|
 Deploy

AI Hub

  • repository to discover, share and deploy ML Models
  • hosted by Google
  • hosts plug and play AI components
  • includes components like
    • Kubeflow Components
    • VM Images
    • Jupyter Notebooks
    • Trained Models
    • Tensorflow Modules

Summary

Product Features Use Cases
AI Building Blocks REST Api Endpoint for vision, language, data Using AI api in apps
Cloud AutoML training models based on custom data Training and deploying models on custom datasets
AI Platform ML pipelines For user to leverage ML pipelines to train and deploy the custom ML models
AI Hub repository for hosting/sharing AI components Resusing existing components