DevIQ enjoyed attending the annual AWS re:Invent conference (something we have been doing for years) in Las Vegas in December 2024. This premier cloud computing event brought together the global cloud computing community for a week of keynotes, technical sessions, product launches, and networking opportunities.
Generative AI aka GenAI was a key focus and there were many opportunities to see how this powerful technology helps modernize and transform business strategy, moving the needle in productivity, capabilities, and cost savings.
Here is a group picture we took of part of our AWS team that attended as well as one of our clients/partners at the final party AWS threw to relax and have some fun.
There is so much content that came out from re:Invent, so DevIQ wants to share the following key highlights that we feel are important:
All About GenAI – a sea change of tools and technologies are transforming how we build software and solve problems. The conference this year was so focused on GenAI that most of our highlights are related to GenAI.
New AI Models and Chips – a new family of Foundational Models called Nova are accessible via Bedrock for text, image, and video generation. Titan nor Nova models will likely compete with top-tier OpenAI and Anthropic models in the near term, yet it clearly establishes AWS as an FM model source that will be very competitive in the long term. Bedrock now supports 50+ Foundational Models to choose from, making model evaluation and selection an important implementation task.
AWS Trainium2 chips are now available for training AI models. And the next generation, Trainium3 is expected to be available by year end.
High-performance RAG – AWS has substantially enhanced its RAG (Retrieval Augmented Generation) capabilities by creating a more integrated ecosystem between Amazon Bedrock Knowledge Bases, ElastiCache, and vector database solutions. The improved architecture allows for more efficient storage and retrieval of embeddings, significantly reducing latency in RAG applications. The integration of these services provides a more streamlined approach to building context-aware AI applications, with built-in support for handling both structured and unstructured data. These improvements make it easier to build and deploy production-grade generative AI applications with better accuracy and reduced hallucinations.
Responsible and Confident AI – AWS Bedrock Guardrails, a tool that helps developers safely configure and build AI applications, requiring AI expertise and leadership to help any organization modernize their SaaS or Application with GenAI technology. These tools help build more trust with those unfamiliar with the care AWS and its partners have already installed to safeguard the use of this transformative technology.
AWS has released new Guardrail features in Amazon Bedrock, including Automated Reasoning checks and multimodal toxicity detection. As an organization, reducing, if not eliminating hallucinations (inaccurate AI responses) is a key focus. These safeguards ensure that organizations can confidently deploy generative AI applications, knowing that the GenAI solution operates within its strict guidelines and limitations.
GenAI Prototyping Tool – AWS has created a simple introductory GenAI tool named PartyRock that allows anyone to conceptualize a GenAI application. The process brings out the important reality that many businesses have yet to find a solid technology partner and explore the powerful new tools/upgrades that could increase productivity, efficiency, revenue, and cost savings. As an example, one DevIQ team member used PartyRock to create a rudimentary GenAI application that effectively estimated out-of-pocket costs for surgical procedures due to a recent experience of waiting 12 hours for an insurance customer service representative to produce an estimate of the costs. The application took only minutes to produce a more accurate estimate!
Multi-Region Active-Active PostgreSQL – Amazon Aurora DSQL represents a significant leap in distributed SQL database capabilities, introducing a serverless architecture that automatically scales across multiple Availability Zones and regions. The AWS developer world has long awaited a cross-region highly available master-master Postgres database solution. This new offering maintains PostgreSQL compatibility while providing enhanced distributed query processing for improved performance at scale. The system automatically handles data distribution and query optimization, eliminating the need for manual sharding or complex application logic. DSQL's ability to maintain ACID compliance while operating in a distributed environment makes it particularly valuable for applications requiring both scalability and strong consistency.
Boosting Lambda Performance – To improve Lambda performance and save cost, engineers should be conscious of how to optimize the Lambda execution lifecycle. As a best practice for minimizing time spent on the INIT execution, developers should:
- Import only what you need and optimize the libraries that you do
- Lazy initialize shared libraries
- Establish connections in the handler function
- Preserve Data and State for subsequent invocations in the execution environment
- Use provisioned concurrency
Lessons in "Simplexity" – AWS's CTO, Dr. Werner Vogels, delivered his annual keynote address to the usual jam-packed room. Vogels shared some valuable insights and wisdom regarding highly scalable cloud engineering practices. Click here to read some of Shawn Davison's (AWS Practice Lead) favorite points from Dr. Vogels' "Lessons in Simplexity".
GenAI Creates Leverage – Not everyone understands that the purpose of Generative AI technology is not to replace jobs/people but to augment, accelerate, and enhance human performance. For instance, mathematicians were not replaced by the invention of the calculator. Instead, the new tool boosted human performance and efficiency, giving mathematicians more time to focus on higher-level functions and problems. GenAI enables leverage, not unlike a wheel, pulley, or wedge that enables physical leverage.
GenAI Can Be Fun – AWS had a GenAI Booth that allowed people to create versions of themselves. Pictured here is Shawn Davison (Managing Partner) who describes the final picture as a “Confucian Scholar who doubles for Keanu Reeves”. LOL.