Doug: Hi, I'm Doug Kelly, machine learning solutions engineer and instructor for the Advanced Solutions Lab in Google Cloud. In this module, we will discuss TensorFlow Extended, Google's production machine learning platform based on TensorFlow. Let's get started with an overview of TFX. TFX is a Google production scale machine learning platform based on the TensorFlow ecosystem, widely used internally at Google and fully open sourced in 2019. It provides a flexible configuration framework and shared libraries to integrate common machine learning tests implemented as components needed to define, launch, and monitor your machine learning system. TFX makes ML ops easier through all phases of the machine learning project lifecycle from prototyping to production. It is designed to orchestrate your machine learning workflow with portability to multiple environments and orchestration frameworks in mind. This includes Apache Airflow, Apache Beam and Coop Flow. TFX already supports four standard deployment targets for TensorFlow models, deployment to TF serving, a high performance production machine learning model server for batch and streaming inference, deployment through a TF light model converter for inference on IoT and mobile devices, deployment to web browsers through TensorFlow JS for low latency web applications, deployment to TF Hub, a model repository for model sharing and transfer learning. It is also portable to different computing platforms, including on-premise and cloud providers, such as Google Cloud. In fact, TFX runs on top of AI platform pipelines, which interoperates with several manage Google Cloud services, such as AI platform training, AI platform prediction, and data flow for distributed data processing. By implementing your TFX workflow on Google Cloud, you can, first, scale your machine learning workflow with your data. TFX simplifies the use of distributed compute resources by leveraging data flow for processing large datasets and computing model evaluation metrics. Second, increase your development and experimentation velocity. TFX enables you to run multiple pipelines in parallel locally or in the cloud backed by AI platform. This includes pipelines with different sets of data splits, models, and hyper parameters to iterate towards an optimal model that you can deploy to production faster. Third, automate your machine learning operational processes for individual and multiple machine learning pipelines. Google Cloud development tools like Cloud Functions, Container Registry, and Cloud Build streamline TFX code sharing, testing, and deployment for continuous training. AI platform pipelines and its integrations with Cloud Storage and Cloud SQL further enable you to regularly retrain, evaluate, and deploy TensorFlow models in production, while keeping tidy ML projects and standardizing artifact and metadata tracking for you. TFX is the most widely used machine learning platform at Alphabet. It is powering tens of thousands of user and programmatic machine learning pipelines at Alphabet subsidiary companies like DeepMind, Verily, and Waymo. They're all doing cutting edge machine learning research on top of TFX with enormous positive social impact applications, such as self-driving cars and continuous monitoring health wearables, Google's 11 core products, including core machine learning components, and Search, Ads, and YouTube. Each of these is serving over a billion plus users. Also Google Cloud's fully managed AI products, such as AutoML Natural Language and autoML Vision, as well as the recommendations API. TFX pipelines underpin all of these services for GCP customers. Since TFX was fully open sourced in 2019, TFX has quickly established an active open source community and corporate partners that have further extended TFX's capabilities and adapted it to improve their own machine learning production workflows. Open Source TFX enables global machine learning use cases, such as Twitter's Cortex machine learning platform, which incorporates TF X to support large scale machine learning applications, such as tweet ranking for its 100 million plus users. Airbus is also using TFX pipelines to automate large scale anomaly detection and report generation for human reviewers of thousands of International Space Station sensors aboard the Columbus scientific laboratory. SAP Concur is using TFX to simplify its Bert natural language model deployments for conversational agents, as well as Yahoo Japan, which is extending TFX to fit its ML needs, and even built an auto ML framework on top of TFX that is serving models for 20 plus consumer facing services, such as article ranking. TFX is the latest evolution in Google's machine learning pipeline infrastructure from over 20 years of doing web scale production machine learning. In 2007, Google first built Sibyl to solve these problems, and it was used widely at Google all the way through 2019. Sibyl was originally intended to be the world's best large scale logistic regression system. However, as Google built a lot more production machine learning, TensorFlow and deep learning became increasingly important, motivating the need for a more flexible solution to support a broader range of machine learning use cases. As a result, the Google Brain team build TFX using everything they've learned about productionizing machine learning over the previous decade. And in 2019, this culminated in open sourcing TFX so that people around the world could use it to build better, more efficient machine learning based software. ML ops is still an emerging engineering discipline and is a highly active area of research and tooling development. Google Research is a recognized industry thought leader in building production machine learning systems and is continuously publishing research on ML best practices that is driving this field forward. TFX represents the culmination of learned best practices around data and model management from over a decade of industrial scale machine learning, serving billions of Alphabet users. By implementing your machine learning workflows with TFX, you are leveraging the machine learning best practices and tooling used at Google to improve your projects probability of success. Running TFX on Google Cloud further supports your project success on Google's infrastructure at scale.