AWS Goodies - June 26, 2024

I am back at my keyboard after spending a fun vacation week in Paris with my wife Carmen Barr . We saw lots of amazing sites, had some great meals, and enjoyed walking around this beautiful and history-filled city.

I have collected plenty of AWS goodies to share and this post may be a bit longer than usual. So here we go!

Amazon Bedrock Video - Before I left for Paris, I recorded a new video to list the Top reasons to build & scale generative AI applications on Amazon Bedrock . After some editing and special effects, the finished video came out great. Check it out and please feel free to share it with your colleagues so that they can learn more about what Amazon Bedrock has to offer:

Building Infrastructure for Generative AI - A new blog post talks about 4 ways AWS is engineering infrastructure to power generative AI . After recapping some of our history building large-scale data centers and GPU-based servers, the post covers low-latency large-scale networking, continuous improvements to energy infrastructure (research indicates that AWS is up to 4.1 times more efficient than on-premises), ground-up security, and our AI chips (Trainium and Inferentia). Here are some fun facts about the upcoming Trainium 2:

Trainium2 is designed to deliver up to 4 times faster training than first-generation Trainium chips and will be able to be deployed in EC2 UltraClusters of up to 100,000 chips, making it possible to train foundation models and large language models in a fraction of the time, while improving energy efficiency up to 2 times.

AWS News Feeds Dashboard - This is really cool. Per the post from Yuriy Prykhodko , the AWS News Feeds Dashboard provides a consolidated view for AWS new feeds including What's New, Blog Posts, Videos, and Security Bulletins. You can view the public demo dashboard or you can follow these steps to deploy it yourself, and you can even view the newest YouTube videos directly from the dashboard!

Let's Build a Startup - My colleague Giuseppe Battista let me know about two recent episodes of the Let's Build a Startup live stream:

• Anatomy of a Unicorn - In this episode they chat with former AWS-er Julien SIMON to learn more about Hugging Face, Generative AI, the importance of Data, and the Open Source community.
• XRAI Glass - In this episode they talk about XRAI Glass and how they are using Amazon Transcribe & Amazon Translate to make the world a better place one word at a time. The post also shows you how to build a simple event-driven serverless architecture that choreographs both of these services:

Serverless RAG - Giuseppe also shared a link to a new blog post, Serverless Retrieval Augmented Generation (RAG) on AWS :

The year is 2024 and you're still paying for a vector database when you're not using it. Not anymore! In this post we explore a fully serverless solution for your Retrieval Augmented Generation (RAG) applications on AWS backed by Amazon Lambda, Amazon Bedrock, Amazon S3, and LanceDB.

The post is detailed and helpful, and includes an in-depth look at the economics of serverless RAG.

State of Amazon Location Service - AWS DevTools Hero Yasunori Kirimoto gave a talk at the AWS Tokyo Summit to review the State of Amazon Location Service :

Intelligent Sharing in Amazon Redshift - My colleague Marc Brooker shared a link to a very detailed, 10-author academic paper (Intelligent Sharing in Amazon Redshift ) that discusses RAIS, an intelligent scaling model for Amazon Redshift. As the summary says:

In this paper, we describe RAIS, the latest collection of AI-powered scaling and optimization techniques in Amazon Redshift, released in preview at re:Invent 2023, which enable it to scale both vertically and horizontally to adapt to all types of workload variability. RAIS dynamically provisions compute resources to run heavy queries efficiently and automatically optimizes warehouse size for the customer’s workload, even as it shifts over time. We show that, depending on the workload, RAIS improves either cost or average query execution time by up to 7.6x and 14.2x, respectively, over existing baselines.

That's about all I have time and space for today!