Beyond Jupyter Notebooks

Build an interactive visualization and filter for a dashboard using Superset, by uploading the data file, configuring the table, and creating a filter chart that refreshes the view.

Configure a docker compose workflow with Jupyter notebook, Superset, and Postgres, mount scripts and volumes, then connect a data frame to a database and visualize results.

Load data from a postcode database, train an arbitrary model, save it as a trend perception file, and deploy the classifier as a prediction service with an object store.

learn to build a RESTful api with API Star to let others interact with your model and return predictions, enabling a prediction-as-a-service api with automated api documentation generation.

Explore API Star, a web API framework, by building an http endpoint with a handler, routes, and an app. Run locally on port 5000 and test with a name parameter.

Build a custom API Star image with docker on a slim python 3 base. Expose port 8000 and map host 5000 to access the API Star endpoint.

Build and customize an AP star docker image to power a prediction service, installing scikit-learn, pandas, and numpy, then run docker build -t AP star:latest to deploy.

Extend your custom wine prediction project by exposing two rest endpoints, predict and retrain, to serve input features and return probabilities, with a secret key and the provided docker file.

Create a secured retrain endpoint that validates a token, loads data, trains and persists the classifier with a timestamp, and expose a predict endpoint returning probabilities as JSON.

Explore how scheduling tasks at fixed intervals enables automated execution, from cron jobs to fetching external data and retraining models to keep predictions up to date.

Explore Apache Airflow, a scheduling framework for orchestrating tasks in workflows, and learn core concepts like tasks, directed graphs, data fetch, feature derivation, storage, and model training.

Launch apache airflow with a community image and run a hello world task via docker compose. Explore the web UI at localhost:1888 and review basic task instances and deck resources.

Explore creating a dag in Airflow by configuring default arguments, setting a 30-minute schedule, and building a PythonOperator-based workflow that logs time, sleeps, and prints hello world.

Leverage Airflow to implement a scheduled model retrain. Fetch the last model, retrain on a random subset of database, and save the updated model with a timestamp using boilerplate infrastructure.

Learn how to build a course project solution by retraining a model, wrapping the retraining function in a Python operator, and scheduling with Airflow.

Master the course content beyond jupyter notebooks, using jupyter for computation, dashboards, storage, and deployment, and explore the opinionated stack with cache, graph database, nginx, Prometheus, and solr.