The last blog in the machine learning series delved into the machine learning capabilities of Microsoft Azure. In this blog, let’s look at the intent behind Google’s AI platform. Google has brought about a true paradigm shift in the way ML models are deployed by completely abstracting the complexity of the underlying algorithmic mechanics. In my opinion, this will work in 80% of the cases with minor tweaks, if any. This is a blessing for organizations struggling with productionizing ML models for their businesses in a meaningful and rapid manner.

ML architecture needs to be flexible to accommodate the elastic learning patterns of the ML process and the large and varying volumes of data and processing power involved. A full-blown enterprise ML architecture, containing all the features above, probably won’t be necessary when just starting out with ML. Early on, practitioners can purchase a small-scale ML platform to suit a specific use case and purchase off-the-shelf tools to support a smaller scale.

Organizations must envision a revitalized data and analytics end-to-end architecture that incorporates diverse data, models, and algorithms and can be accessed anywhere, if they wish to support ML applications.

General Architecture

Google’s AI Platform is completely GUI driven and requires minimal coding or statistical experience. The platform democratizes AI using Deep Learning, making it fast, accessible, and easy for enterprise-wide deployment. With the Cloud AI platform, you can train, evaluate, improve, and deploy models through a graphical UI. This simply involves uploading data and following the UI-driven instructions to select the appropriate instances /clusters based on the performance requirements and data size.

Here is a visual representation of the complete modeling lifecycle within Google’s AI Platform:

Modeling lifecycle within Google’s AI Platform

Unique Features

The unique aspects of Google’s AI Platform worth mentioning are AI-Hub and Kubeflow Pipelines.

The AI Hub is a one-stop destination for plug-and-play ML content, including pipelines, Jupyter notebooks, TensorFlow modules, and more. It offers two significant benefits. The first is making high-quality ML resources developed by Google Cloud AI, Google Research, and other teams across Google publicly available to all businesses. The second is that it provides a private, secure hub where enterprises can upload and share ML resources within their own organizations.

Kubeflow Pipelines are a new component of Kubeflow, a popular open-source project by Google that packages ML code just like building an app so that users across an organization can reuse it. Kubeflow Pipelines provide a workbench to compose, deploy, and manage reusable end-to-end machine learning workflows, making it a zero lock-in hybrid solution from prototyping to production.

Performance and Benchmarks

Kaggle Leaderboard delivers superior performance despite complete automation, with accuracy at the rear end of the Elbow curve. Google’s AI Platform is a ready-to-deploy, highly abstracted model with minimal tuning requirements, and its results can be seen below.

Kaggle Leaderboard 

How AutoML Tables Performs 

Other Applications Supported by Google

Google offers a rich ecosystem of AI products and solutions, ranging from hardware (Tensor Processing Unit [TPU]) and crowdsourcing (Kaggle) to world-class ML components for processing unstructured data like images, video, and text. Google is also one of the pioneers of automated ML with its Cloud AI Platform.

Google provides AI services on two levels: a machine learning engine for savvy data scientists and a highly automated Google Prediction API. Google does not disclose which algorithms are utilized for drawing predictions and does not permit model customization.

Text & Speech

Google has a unique API in terms of Dialog Flow to analyze intent that is comparable to other competitive platforms.


The tool built by Google for image recognition tasks is quite powerful for finding specific image attributes like labeling objects, detecting faces, analyzing expressions, finding landmarks, describing scenes, spotting text in images, and identifying dominant colors.


Google supports various image processing tasks but lacks video analysis features provided by Microsoft and Amazon.


ML Platforms, including Google’s AI Platform, are generally charged based on hourly usage, and language apps are charged by the number of words/characters.

Typical Use Cases

Fox Sports embarked on a one-of-its-kind journey with the AI Platform Tables application by training the model on multiple variables of historical cricket matches which could predict when a wicket would fall 5-minutes before it happened on the pitch. Monty & AI Platform Tables changed the game forever with user engagement up 140% versus industry averages and the marketing for this activity delivering 150% more subscribers per dollar spent

Concluding Remarks

The successful deployment of ML initiatives will require specialized skills, a few of which will likely already be present in the organization. Google’s AI Platform has minimized this requirement to a large extent, and organizations can also benefit from strategic partners for any sort of customizations rather than building this capability from scratch. We at Persistent are one such strategic partner, helping organizations realize the dream of machine learning in their setups.


Tackling High-Value Business Problems Using AI Platform on Structured Data (Cloud Next ’19)