KubeFlow Bootcamp

Gain a quick overview of Google Cloud basics for Kubeflow, including cloud networks, advantages, account setup with a $300 credit, and a budget alert demo.

Define cloud computing with the five NIST traits and trace its evolution from time sharing to virtualization to serverless, highlighting Google Cloud's container-based serverless approach.

Explore how Google Cloud network infrastructure connects computers worldwide through regions, zones, and multi-region locations, enabling fast storage and compute with fault tolerance.

Learn how Google Cloud network connections route user requests through edge points of presence and edge nodes, using peering and CDN caches, to reach data centers for services like BigQuery.

Discover why choose Google Cloud Platform for compute, storage, big data, and machine learning, with open accessibility, hybrid cloud options, robust security, and clear budgeting tools.

Learn to set up a Google Cloud account, claim $300 free credits, and tour the cloud console to manage projects, services, and a shell for KubeFlow bootcamp demos.

Learn to manage Google Cloud billing and budgets using billing reports, cost tables, and alerts to monitor forecasted costs across projects, with automated actions via Cloud Monitoring and Pub/Sub.

Take a guided tour of Google Cloud billing tabs, review credits and cost views, and set up your first budget to manage costs before using resources.

Set up and manage a Google Cloud budget by creating a monthly budget with scope to all projects, applying credits, and configuring thresholds including a 1% alert and forecasted spending.

Survey Google Cloud storage options, including object storage with Cloud Storage and non-relational options Firestore and Bigtable, plus relational databases Cloud SQL and Cloud Spanner.

Explore Google cloud storage concepts, including buckets and objects, storage classes from standard to archive, access controls, versioning, lifecycle policies, and transfer options for scalable, public or private data.

Learn how to create a cloud storage bucket, upload and download objects, share a publicly accessible link, organize with folders, and clean up by deleting objects and the bucket.

Understand how Kubernetes underpins Kubeflow, from basics: containers, nodes and pods, control plane, Kubernetes API, to advanced topics like kubectl, deployments, deployment strategies, pod networking, storage with demos.

Learn the basics of Kubernetes, a portable open source container orchestration platform that uses declarative configuration and automation to manage microservices, containers, pods, and clusters at scale.

Explore how containers package apps with isolation, compare containers to virtual machines, and use docker and kubernetes to build, deploy, and manage pods representing your microservices.

Discover how Kubernetes organizes nodes and pods with the control plane and Kubernetes API, and how API server, etcd, scheduler, and the controller manager coordinate containers across the cluster.

Learn how to use the Kubernetes API to deploy a hello world app, compare imperative and declarative workflows, and manage deployments, pods, and services with kubectl and gcloud.

Explore how container images use Dockerfile layers and a writable top layer to run multiple containers on one kernel, with cloud Build and Gcr.io on Google Cloud.

Set up Cloud Build and Google Container Registry, create a hello world shell script and Dockerfile, and deploy using a declarative cloudbuild.yaml with history in the registry.

Explore Google Kubernetes Engine, a fully managed Google Cloud service that deploys, manages, and scales Kubernetes with standard and autopilot modes for containerized applications.

Explore the Kubernetes architecture, including the control plane, API server, etcd, scheduler, and controller manager, and how kubelet, kube-proxy, and kubectl interact on regional versus zonal GKE.

Learn to create a GKE cluster via the Google Cloud Console GUI, deploy a sample pod, and manage node pools and workloads; compare GUI deployment with kubectl and gcloud CLI.

Learn how kubectl connects to the Kubernetes API server in the control plane to manage resources, using credentials and config, with commands like get pods and flags to customize output.

Demonstrates configuring gcloud for gke, creating a cluster, deploying a hello world container from artifact registry, exposing via load balancer, and validating access via the external IP.

Learn how Kubernetes deployments translate a declared desired state into running pods, manage replica sets, and perform rolling updates and rollbacks for stateless apps using declarative, imperative, or graphical methods.

Learn how to update deployments with rolling updates that avoid downtime, using kubectl apply, kubectl set, or kubectl edit to adjust image, replicas, and resources.

Explore Kubernetes deployment strategies, including blue-green and canary, using services, labels, and traffic splitting to update apps with minimal downtime, and manage rollouts with kubectl.

Demonstrates deploying nginx with a Kubernetes deployment and triggering rollouts via kubectl, including a deployment manifest and a two-node GKE cluster setup in Cloud Shell.

Understand Kubernetes pod networking and the IP per pod model, where each pod has an IP in a shared network namespace, with alias IP ranges supporting services in GKE.

Explore Kubernetes storage with volumes and persistent volumes, including ephemeral emptyDir, configMap, and secret, downward API, and how PVCs bind to PVs through storage classes such as standard and SSD.

Demonstrates creating a persistent volume claim on Kubernetes in GKE and deploying an Nginx pod that uses the PVC for persistent storage.

Explore machine learning fundamentals and how Kubeflow integrates with Kubernetes. Apply supervised learning, ML algorithms, and Google Cloud tools, including Vertex AI workbench notebooks.

Explore the machine learning pathway from real-world data to data products by collecting, cleaning, organizing, and exploring data, then applying supervised or unsupervised learning to build dashboards and services.

Discover why machine learning uses data to automatically infer important features and predict labels, with examples like credit scoring, price forecasting, and spam filtering, and why data quality matters.

Learn how supervised and unsupervised learning differ, contrasting historical labeled data with unlabeled data, and introducing classification and regression tasks, with a path to deeper exploration of the supervised process.

Master the supervised learning process: from defining features and labels to train-test splits, model training, and evaluating predictions with hyperparameter adjustments and cross-validation.

Compare artificial intelligence, machine learning, and deep learning, distinguishing narrow vs general AI, and learn how vertex AI and deep learning models tackle classification, image, NLP, and speech tasks.

Explore Google Cloud AI and ML options, from custom model coding to AutoML and pre-trained API services, under Vertex AI.

Explore Vertex AI, a unified Google Cloud platform that builds, deploys, and scales ML models from data prep to production, with AutoML, notebooks, and model monitoring.

Launch and manage Vertex AI Workbench notebooks on Google Cloud to run Jupyter labs, configuring CPU or GPU, and using Python libraries like pandas, scikit-learn, and TensorFlow.

Explore the Kubeflow ecosystem and its strong link to Google Cloud, including Vertex AI Pipelines, Kubeflow components, and the Kubeflow Python SDK, with hands-on demos.

Explore Vertex AI pipelines to automate, monitor, and govern ML workflows with a serverless, Kubeflow-backed platform, using Vertex ML metadata to capture lineage from data ingestion to deployment.

Compare Kubeflow pipelines SDK with TFX and learn how Vertex AI pipelines orchestrate scalable, serverless ML workflows on Google Cloud.

Demonstrate Vertex AI pipelines with Kubeflow integration inside a Google Cloud notebook, install KFP, build components, compile and run pipelines, and monitor artifacts in Google Cloud Storage.

Explore Vertex AI pipelines by running a Kubeflow pipeline from a notebook, including setting up a storage bucket, enabling APIs, and compiling a simple emoji-based pipeline.

Discover Kubeflow architecture, from pipelines and components to accelerators and notebooks, running on Kubernetes, and navigate experimental and production phases covering data prep, training, evaluation, and model serving with monitoring.

Define Python components with the Kubeflow pipelines DSL and decorators, then assemble them into a directed acyclic graph pipeline on Kubernetes and compile to YAML for deployment.

Explore Kubeflow pre-built components that connect pipelines to Google Cloud, AWS, and Azure services using KFP DSL. Learn to locate, use, and customize components via GitHub and Vertex AI documentation.

Explore lightweight python components in Kubeflow pipelines, using the dsl component decorator with type-annotated inputs and outputs. Manage dependencies and base images inside the function for runtime containers.

Learn to convert lightweight Python components into containerized components for Kubeflow pipelines by structuring src directory with helper modules, updating the component decorator, and building a container image for reuse.

Compile Kubeflow pipelines to YAML with the PHP SDK compiler to produce a hermetic pipeline representation and read the top-of-file comments for name, description, inputs, and outputs.

Demonstrates building a Kubeflow pipeline on Google Cloud, using BigQuery data, training a scikit-learn model, and deploying the endpoint via Vertex AI; includes setup, compilation, and debugging tips.

Discover how Kubeflow orchestrates machine learning pipelines to enable continuous training, integration, and delivery. Automate data collection, training, evaluation, and deployment with Vertex AI pipelines, Cloud Build, and triggers.