Berlin AWS UG virtual meetup

----------
Agenda:
19:00 - "Time for an opinion: Serverless Framework Components"

International speaker: Gareth McCumskey (serverless.com)

Speaker bio: "Been building web sites since the early 2000's and started my first professional web development role in 2008. In 2015 got into building web applications using the Serverless Framework and it has literally changed my life. Serverless Inc noticed my enthusiasm and asked me to join them early 2019.
Work from home in Cape Town, South Africa with my wife and two kids."

Abstract:
When the Serverless Framework released back in 2015, it quickly became and remained the most popular developer framework for Serverless Application development. Things have slowly been changing in the developer eco-system however, and the non-opinionated way of composing an application of cloud based managed services has shown its short comings, primarily confusion by developers as to how to build an application using Serverless technologies.

Enter components; a collection of opinionated Serverless outcomes that take the minimum of input to deploy into the cloud and allows developers to improve not just the capability to launch their Serverless applications fast, but improves developer experience in many ways!

Level: 200

19:45 - "Continuous Delivery for AutoML - CD4AutoML" -- Automate the end-to-end AutoML lifecycle with Amazon SageMaker Autopilot and Amazon Step Functions

Local speaker: Olalekan Elesin (HRS Group)

Speaker bio: "An AWS Machine Learning Hero, Lekan is an engineer at heart with strong affection solving problems with technology.
He has vast experience with building data and machine learning infrastructure on Amazon Web Services."

Abstract:
The growing popularity of AutoML is fueled by the promise of putting machine learning in the hands of every engineer. However, the majority of AutoML projects do not go beyond Jupyter Notebooks or personal computers. As new data arrives on a consistent basis, AutoML models need to be retrained for model freshness and better predictions.

To achieve this, it is imperative to have an automated deployment process from AutoML model training to serving predictions to our customers.

In this talk, we explore how to train a model with Amazon SageMaker Autopilot in Amazon SageMaker Studio and visualize training experiment results. After this, we will delve into an end-to-end machine learning workflow with Amazon SageMaker Autopilot orchestrated with Amazon Step Functions. This workflow will cover the following:

• Retraining our Amazon SageMaker Autopilot model on schedule
• Retraining upon arrival of new data in Amazon S3
• Automatically deploy a serverless REST API with Amazon Lambda and Amazon API Gateway to serve real time predictions in any application.

Finally, we will assess how Continuous Delivery for AutoML (CD4AutoML) in practice helps to bring the benefits of Continuous Delivery principles to Automated Machine Learning (AutoML) Applications.

Level: 200