1. SageMaker Unified Studio
The biggest wonder of all is the re-designed and re-architected SageMaker to be not just for AI but for a convergence of data, analytics and GenAI. The old SageMaker for AI is now called appropriately ‘SageMaker AI’. The new SageMaker Unified Studio is a task-oriented tool for my data projects with an end-to-end workflow that includes:
Amazon Bedrock IDE
Amazon SageMaker AI
Integration with AWS Data services
Amazon SageMaker Lakehouse
Jupyter-style Notebooks are more streamlined for a smooth end-to-end workflow.
Announcing the preview of Amazon SageMaker Unified Studio
2. Nova Foundation Models
Amazon’s surprise of a set of new Foundation Models is not just low-cost but competes strongly with Gemini and OpenAI and Claude models. And it’s 75% off low-cost as well. It’s optimized for the enterprise and very customizable.
Amazon Nova Micro = Text only. Lowest latency. Low cost. Can be fine-tuned and be a student model for distillation.
Amazon Nova Lite = Low-cost and fast. Can be a student model for distillation.
Amazon Nova Pro = Best combination of accuracy, speed and cost. The standard model where the other models are big and small and this one is regular. Can be a teacher model in distillation.
Amazon Nova Premier = This model is the best at reasoning. Can be a teacher model in distillation.
Amazon Nova Canvas = Generates high quality, studio images.
Amazon Nova Reel = Generates very professional short videos for business.
3. Amazon Q Developer
Now an end-to-end tool for building AWS solutions. Code reviews, documentation & testing. Bootstrap new projects with a single prompt, generate unit tests with a single prompt, create enhanced documentation with a single prompt, create code reviews with a single prompt to detect and resolve security and quality issues. This end-to-end workflow is throughout SageMaker Unified Studio as an assistant with a free tier included.
4. AWS Security Incident Response
This deals with security findings of Amazon GuardDuty and other third party tools within the AWS Security Hub. It provides access to AWS Security experts.
New AWS Security Incident Response helps organizations respond to and recover from security events
5. Amazon Connect
When I need a Cloud Contact Center, I want one that innovates the customer experience. Amazon Connect has all the bells and whistles that become essential features: personalization, automatic recommendations, tracking custom issues with multiple interactions, etc. With AWS Contact Lens I can get advanced AI and ML capabilities in the customer experience. It now has better, more streamlines agents and omnichanell WhatsApp business messaging.
Newly enhanced Amazon Connect adds generative AI, WhatsApp Business, and secure data collection
6. Trainium3 AI Chip
I’ll use the AWS Trainium AI chip if I’m building a big AI model from scratch. I can use if for both training and inference. But if I’m doing inference, I can just use the Inferentia chip for low-cost and also energy efficiency (if I have a green thumb.) However, Trainium will be more powerful.
Trainium3 coming in late 2025 is four times more powerful than Trainium2. It’s so powerful that AWS will need special cooling tech in their data centers.
Amazon EC2 Trn2 Instances and Trn2 UltraServers for AI/ML training and inference are now available
7. S3 Tables and S3 Queryable Object Data
S3 Tables are in Apache Iceberg format, a popular way to handle files in parquet format. Compared to creating your own tables, it’s three times faster for queries and ten times faster for transactions. I’ll use it for daily purchase transactions, streaming sensor data, ad impression, etc.
There can be so many object piling up in S3 that object metadata becomes vitally important. I may want to quickly find data for analytics, data processing or my AI training workloads. When stored in S3 tables, I’ll get automatic generation of metadata with over 20 elements. There’s a new Metadata tab in the Amazon S3 console.
New Amazon S3 Tables: Storage optimized for analytics workloads
Introducing queryable object metadata for Amazon S3 buckets (preview)
And those were the Seven Wonders of AWS re:Invent 2024. Many of them were about AI and even if they weren’t they certainly will have AI behind the scenes in some way. Let the AI revolution continue and we’ll see what wonders we ourselves can create when exploring the new tools at our disposal. These are tech wonders that enable us to create our own.