A new direction for my career
For the past seven years I’ve been working with builders who want to deploy containers on AWS. You may know me from personal assistance I’ve given you in direct messages, from reading my social media posts and blog posts about Amazon ECS and AWS Fargate, or from using my reference architecture patterns on Containers on AWS. Today I’m announcing a change of role inside AWS. I will be focusing on generative AI services, in specific Amazon CodeWhisperer and Amazon Q with CodeWhisperer.
Prior to joining AWS I was a software engineer in the NYC startup scene. I could write the code, but I needed a better way to get that code running in production. Containers captured my attention when I saw what they were enabling. With container tools even a small startup could have highly efficient but highly resilient application deployments. Startups using containers were unlocking a new way to think about infrastructure that was previously only common for the largest companies in the world, and even then only if those companies built custom deployment orchestration software maintained by large internal teams.
Just like containers did, all good technologies enable builders to do greater and greater things with fewer resources. The cloud as a whole is another technology that I love because of how it enabled startups. Amazon CloudFront gives me 1TB of free data transfer out to my users. DynamoDB gives me 25 GB of free database storage, and up to 200 million free API calls to access that data. AWS Lambda gives me 1 million free requests per month. I can run a decent sized application for almost nothing. And if I’m willing to spend $100 a month I can maintain a really large business, with multiple copies of the application and it’s data, distributed across redundant hardware in multiple data centers, powered by renewable energy sources, connected by redundant network capacity.
If I have an idea that has any form of product market fit at all, then it has never been easier to turn an idea into reality… as long as I have the skills to do so. Going from idea to reality with powerful tools requires powerful skills. All the tools can be there, but if I don’t know how to use the tools, or I don’t know how to use the tools efficiently, then I don’t benefit from the tools.
As Amazon CEO Andy Jassy once famously said: “there is no compression algorithm for experience”. It takes a human a long time to achieve the skills to effectively use complex tools like the cloud. But in recent times we have something new, and it gives us a pretty good approximation of such a “compression algorithm for experience”.
It turns out that you can put an extremely large amount of hardware to work, analyzing human data for the combined equivalent of thousands of years of training time. The result is a Large Language Model (LLM) which exhibits the fascinating capability to produce novel results that were formerly only achievable by a human. Some call it AI. Some call it “fancy autocomplete”. But what I call it is a new tool that we didn’t have before, and it is a tool that can help us use all of our other existing tools better and more efficiently.
Just as I was excited back when I first learned about the cloud, or when I first learned about containers, I am now excited about the possibilities of combining human effort with machine assistance. In the coming months you’ll see me talking more about the generative AI tools that AWS is providing to assist builders. I’ve been using generative AI tools daily for some time, and will continue to do so. I will share more about my experience building with generative AI, and I am eager to learn more about your experiences building with generative AI as well.
More to come soon! But in the meantime if you have questions or concerns you can reach me on a variety of social media websites including LinkedIn /in/nathankpeck.
PS: As a closing note I want to mention Amazon ECS. I believe very much in this product, and it is not without a feeling of sadness that I shift my focus away from serverless container orchestration on AWS. I still believe that the very best way to deploy applications on AWS is in a container under the management of an efficient serverless control plane. I shudder to think of all the wasted compute resources and carbon footprint caused by many different companies each running their own independent Kubernetes control planes. Container orchestration is a prime example of “undifferentiated heavy lifting” that should be an efficient, shared service pool that is utilized by many people. If you’ve followed me and interacted with me primarily for my Amazon ECS focused content, then rest assured that I still believe in Amazon ECS. I may not be able to provide personalized assistance if you have questions or feedback relating to Amazon ECS, but my learning content is still out there, and I will still attempt to redirect you to someone in the team who can assist you as you need.