AWS Summit Amsterdam
AWS is committed to train 25M people in technical skills by 2025.
Sustainability
AWS wants to be Net-zero by 2040. They will use 100% renewable energy by 2025 and are aiming to be water positive by 2030. They've changed to using plant based oil to power backup generators instead of Diesel. By moving workloads into AWS, this reduces your energy consumption by up to 80%.
WeTransfer
Global platofrm with over 80 million users. By using AWS, they've reduced their carbon footprint by 78% while still scaling their business.
Innovation
Lonely Plaet leverages Amazon Bedrock to generate personalized itineraries and reduce generation times by up to 80%.
Picnic
They want to improve sustainability in the food market and reduce waste. For a long time, sustainable and affordable didn't really go together. Picnic is the online supermarket who wants to use the latest technology, but no build on top of it. Take a technology from the past and innovate on top of it. Step 1 is to take all friction away, how do you make it easy to shop online? To deliver a great service to customers, Picnic optimizes the process to buy any product from Europe and deliver it to the customers. Each customer has their own personlized store, generated by AI. The store includes a personalized assortment with local products. Their fullfillment centers are highly automated, optimized for every product.
Generative AI Stack
The bottom layer has infrastructure for FM training and inference. Training takes lots of data and builds a model from that, inference uses that model to generate output. This layer contains services like AWS SageMaker. These tools require a lot of processing, so AWS provides hardware to allow this with NVIDIA GPU enabled instance types.
Project Ceiba is the largest supercomputer in the world and a collaboration between AWS and NVIDIA. AWS built custom chips for this on their own sillicon. AWS Trainium and Inferentia reduce energy requirements to power training and inference.
The middle layer contains tools to build and scale FMs and LLMs. Amazon Bedrock is a fully managed service that provides base models. Bedrock includes guard rails to use AI in a responsible way.
The top of the stack consists of apps that leverage FMs and LLMs. Amazon Q is the genrative AI powered assistant, that's tailored to your business.
PostNL
PostNL delivers 1.1M parcels per day. Over the centuries they've adapted from horeses to cars, from physical offices to online offices and from letters to parcels. So how do they approach software? Before 2012, they mostly used commercial software in their private data centers. In 2012, they discovered that maintaining data centers is not their core busiuness and they moved everything to AWS. In 2018, they started to build their logistics software in house. The goal is to build the services the customers need and that the services are aleays available. Reliability is one of their highest priorities.
PostNL expieriences fluctuating load on a daily basis. By choosing serverless, the development teams do not need to think about seasonality, reducing operational overhead. When using truly serverless services, like DynamoDB and SQS, the maintenance is completely managed by AWS. When Log4j vulnerabilities were found, AWS patched their services to protect their customers. By leveraging serverless, you can delegate the hardest problems to AWS.
Engineers at PostNL are expected to understand distributed systems. This requires long term investment in people and culture. PostNL provides training by their own developers and AWS. An effective way to provide serverless engineering, is to make the right thing easy. They provide one-click setup of new AWS accounts that are automatilly onboarded on their network. Making wrong things hard using guard rails will also help your developers.