Generated by FLUX.1 [Dev] by Black-Forest-Labs
AWS re: Invent 2024 Recap
A month has gone by already, and I am sitting at my desk finally distilling the whirlwind of information that comes from a great conference like AWS re:Invent. Rather than revisit widely discussed announcements, because I know there are plenty of great authors who have already covered those subjects, I’ll share the key trends and practical lessons I gleaned from the sessions and chalk talks I attended.
First Big Trend: Generative AI (GenAI)
Bet you cannot guess it — Generative AI (GenAI). I don’t think it comes as much of a shock to you based on the title graphics, but I am a big fan of generative artificial intelligence. It has been improving at an astounding rate, and there are many use cases for it in various industries to help augment how we currently work. Think of it as a more advanced autocomplete since it is really just guessing at the next sequence of tokens that will best complete your prompt.
Foundation Models
Amazon Bedrock is the highlighted service in this domain. It helps users of this service start developing with GenAI within minutes. It is an Amazon-managed serverless service that gives you access to leading AI companies’ Foundation Models (FMs). You can test which model best fits your use case and then continue to privately customize it with fine-tuning or using Retrieval Augmented Generation (RAG). To add to this RAG technique, Bedrock offers Knowledge Bases. It is better to read how AWS describes this product:
Knowledge Bases for Amazon Bedrock automates the complete RAG workflow, including ingestion, retrieval, prompt augmentation, and citations, removing the need for you to write custom code to integrate data sources and manage queries.
Worried about what an AI agent might say to a customer? Guardrails for Amazon Bedrock can help you establish predefined answers or omissions if you want to guard a user from trying to abuse the AI agent.
One of the niceties that Bedrock provides is the wide range of FMs from AI21 Labs, Anthropic, Cohere, Meta, Mistral AI, Stability AI, and more, all available to interact with under a simple API.
Serverless Approach
Simple Serverless GenAI application
Remember when I said Bedrock was serverless a minute ago? When you combine that with all the other fun serverless services from AWS, you can create a pretty easily managed infrastructure with GenAI capabilities. I linked the same workshop I participated in down in the references if you would like to try it yourself. The UI is served via S3 behind a CDN, while the API Gateway takes the request from the client and forwards it to the appropriate Lambda function. This function contains our HTTP handler and business logic to convert the prompt into the desired form when interacting with our FM. Then the response can be sent back to the client as a payload or via stream if we wanted something similar to a chatbot experience.
Second Trend: Platform Engineering
This could be tied solely to the path I carved at re:Invent, but I was surprised to see how many sessions were talking about Platform Engineering. A lot of the higher-level technical talks I attended discussed the ability to support standardized development without needing to sacrifice the autonomy of your teams. The below graphic points toward the trend of more autonomy meaning harder-to-control standardization and vice versa. Platform Engineering teams try to solve this dilemma by providing standardized templates and tools for teams to self-service from.
How Platform Engineering can help achieve both Autonomy with Standardization
This, of course, was a journey to get to the point of the Platform Engineering team’s role in a company. Traditional Ops had the role of hardware gurus and change management sticklers, while developers would compile programs to try and solve problems for the business or communities.
Then we started introducing the shift-left movement, where Agile and DevOps began to shine a light on innovative techniques for shipping a product and its changes quickly. “asCode” solutions for infrastructure, integration, configuration, and policy (IaC, CasC, PAC) started taking the scene as they enabled developers to quickly configure how their apps should be running and delivered. The caveat to this approach was the fact that all these new tools were extra load on the developer to learn and maintain, and the management of the teams did not know how to standardize the tooling that the teams used.
The key to this was the development of Platform Engineering teams. Teams of software engineers and architects help build and approve blueprints of “asCode” solutions and boilerplate projects that meet the standards of their organization’s needs. These allow the application teams to choose the right tools for the right job. All the previous overhead is abstracted into a self-serve system with a single pane of glass for management to easily control and evaluate the health of their teams.
Evolution of “Shift Left” to Platform Engineering
GitOps
Now, the DevOps movement was covered pretty quickly in the last segment, and this topic is a derivative of that — GitOps. GitOps is still catching on in the industry as it shows a great way of controlling the state of your environments, whether it is runtimes, orchestrators, infrastructure, or cloud providers. The Git in GitOps refers to the technology behind the popular version control system, where one can make changes to the file, and it keeps a history of the changes in a version control system. GitOps is the practice of keeping the whole state recorded and mutable at the fingertips of the developer. All one needs to do is commit and push to the remote, where some program would react to the change and reconcile it. My favorite GitOps toolset comes from the Argo Project!
Backstage
The Platform Engineering team’s main deliverable is an app referred to as the Internal Development Portal (IDP), which many organizations can create from scratch. It automates the process of interacting with other tools that the organization uses. One open-source project under the CNCF attempts to provide the community with a framework to build their own development portals — Backstage by Spotify. Backstage is a React app that has a vast amount of plugins to fit your needs and is extensible so you can make the application look and feel like it was born out of your company’s creative team. Backstage was a popular technology mentioned by big-name companies from HelloFresh to Toyota. They explained how it fast-tracked solutions to empower their teams and increased the ease of onboarding new engineers to the company.
GenAI Agents
Toyota talked about the new way they are using GenAI agents in Backstage to help with tasks like consulting on solutions and even debugging the deployed solutions. The agent in Chaufer (their Backstage app’s name) uses the tech docs and catalogs as context through Retrieval Augmented Generation techniques. This allows the agent to know specifics about the software being used and current deployments. Developers can use it as a handy search for “What’s the best software template to use to build a three-tier web application?” and Chaufer will respond with suggestions and links to the templates hosted in Backstage.
Blueprints
After someone was explaining the concept of Platform Engineering being a blueprint team, it clicked on how I can easily explain the benefit of a team such as this. The blueprints are the software templates that other teams use. When you use tooling like Terraform, Cookiecutter, and Argo, it allows you to streamline ideas to deployed concepts faster than developers could before. You can have apps and projects integrated into your ecosystems from day one. This allows experimentation and faster innovations as you enable your developers the autonomy to create features and keep them happily challenged while providing standardized quality.
Closing Thoughts
AWS re:Invent 2024 reiterated the importance of developer experience and operational excellence in driving innovation. GenAI is no longer a novelty but a critical enabler of transformation, from content creation to real-time analytics. Meanwhile, Platform Engineering paves the way for frictionless collaboration, aligning business needs with technical execution.
As we look to the future, the challenge isn’t just adopting these trends — it’s leveraging them strategically. Are your teams ready to embrace these innovations to not only stay competitive but to lead in an era defined by agility and intelligent systems?
References
- My Own Notes
These are some of the sessions and labs I went to, chalk talks are not recorded:
Top comments (0)