AWS re:Invent 2024 Key Takeaways

Julian Wiffen, Chief of AI and Data Science at Matillion, recently attended AWS re:Invent 2024 in Las Vegas. Let's hear his takeaways from this year’s event! 

Attending AWS re:Invent 2024 in person this year was inspiring if tiring. My iPhone showed me averaging 15,000 steps a day, which speaks to the sheer scale of this event. We (Matillion) had many fruitful and interesting discussions with customers and partners but it was vital to allow plenty of time to find and reach each meeting location! The number of attendees (65,000+) and the sprawling event was matched only by the sheer number of announcements of new products and features for the GenAI space. 

The event was buzzing with energy and innovation, especially around generative AI, and showcased its huge potential to transform how we work with data. For Matillion customers, the announcements underscored immense opportunities to enhance data integration and transformation pipelines, particularly when working with unstructured data and leveraging the power of AI. Let’s dive into the AWS updates that stood out and explore how they align with and amplify Matillion’s capabilities.

More details on each of the announcements can be found here - About Amazon

Amazon Nova: A New Generation of Foundation Models

At re:Invent 2024, AWS introduced Amazon Nova, a suite of state-of-the-art foundation models designed to process text, images, and videos. Available through Amazon Bedrock, the Nova family includes models such as Nova Micro, Lite, Pro, and the forthcoming Premier, each tailored to specific performance and cost requirements.

Amazon Web Services

Why it’s exciting: For Matillion users, integrating Amazon Nova models can significantly enhance data transformation workflows, especially when dealing with unstructured data. The multimodal capabilities of Nova models enable more comprehensive data analysis and content generation, aligning seamlessly with Matillion’s AI-driven data integration tools. We are especially interested in how Nova handles any format to any format - whether it be image to text, text to video, or many other combinations. Nova is also significantly cheaper than other models, opening up the potential for AI use cases on a much larger scale.

Amazon Bedrock Marketplace: Expanding Model Accessibility

AWS unveiled the Amazon Bedrock Marketplace, providing access to over 100 foundation models from leading AI companies. This expansion offers Matillion users a diverse selection of models tailored to specific generative AI use cases, enhancing the flexibility and capability of data transformation workflows.

Why it’s exciting: The Marketplace will enable Matillion customers to seamlessly integrate a variety of AI models into their data pipelines, fostering innovation and allowing for more customized and effective data processing solutions. We look forward to seeing how specialised small language models perform relative to more generalised larger ones. 

Amazon Bedrock Knowledge Bases: A Game-Changer for Retrieval-Augmented Generation (RAG)

AWS’s introduction of Knowledge Bases within Amazon Bedrock opens the door for seamless integration with Matillion’s Retrieval-Augmented Generation workflows. Matillion already enables users to load data into vector stores like Pinecone, which allows large language models (LLMs) to access structured and unstructured datasets. Amazon Bedrock’s Knowledge Bases could amplify this capability by offering:

• Richer Contextual Data: With the automated generation of knowledge graphs and SQL queries, Matillion users can leverage Bedrock to make their pipelines even smarter.
• Streamlined Workflows: The combination of Bedrock’s robust RAG capabilities with Matillion’s AI Prompt components ensures that data teams can integrate high-quality, AI-enriched insights directly into their workflows.

Why it’s exciting: For customers tackling messy or unstructured data, this partnership could redefine what’s possible, enabling smarter data enrichment and more intuitive analysis without custom coding. Early experiments with knowledge graphs powering graph-RAG are already showing improvements in the accuracy of responses to complex technical questions. 

Amazon Bedrock Data Automation: Transforming Unstructured Data

AWS’s Bedrock Data Automation tool introduces a generative AI-powered ETL framework designed to handle multimodal content such as PDFs, audio, and videos. This aligns perfectly with Matillion’s vision of empowering data teams to transform and prepare unstructured data efficiently.

Matillion’s low-code interface and AI-enhanced pipeline building already make it easy for teams to integrate unstructured data. Bedrock Data Automation can take this a step further by:

• Simplifying Data Preparation: Automating the extraction, transformation, and loading of non-tabular data.
• Expanding Use Cases: Enhancing AI-driven capabilities for industries like healthcare, finance, and media, where unstructured data often reigns.

Why it’s exciting: This could unlock the full potential of multimodal data at scale by allowing a much wider range of formats to be brought into a Matillion pipeline.

Enhanced Prompt Management and Multi-Agent Collaboration

AWS’s advancements in prompt management and multi-agent collaboration through Bedrock resonate strongly with Matillion’s AI Copilot and Prompt Components. These tools offer:

• Better Workflow Coordination: Multi-agent orchestration in Bedrock mirrors Matillion’s ability to build complex, scalable pipelines with minimal effort.
• Prompt Caching: Cache frequently used context in prompts to reduce latency - highly relevant when processing large batches of data
• Prompt Routing: Enabling prompts to be directed to a range of models based on their complexity
• Optimized AI Prompts: Advanced prompt handling ensures efficient communication with LLMs, further enhancing the AI-driven tasks in Matillion pipelines, such as data classification and enrichment.

Why it’s exciting: Customers will be able to execute more sophisticated AI workflows with greater efficiency, driving better results across diverse use cases.

Amazon Bedrock Model Evaluation: Ensuring AI Accuracy

AWS introduced new model evaluation capabilities within Amazon Bedrock, including the LLM-as-a-judge feature for assessing model performance and Retrieval-Augmented Generation (RAG) evaluation for Knowledge Bases. These tools assist in selecting the most suitable models for specific tasks, ensuring that AI applications deliver accurate and reliable results.

Why it’s exciting: For Matillion users, these evaluation tools provide confidence in integrating AI models into data transformation processes, ensuring that outputs meet quality standards and align with business objectives. More excitingly, they offer the potential to speed iterative improvements to GenAI pipelines by allowing new models and other variations to be tested quickly as opposed to the currently slow and manual evaluation process. 

Partnership with Poolside: Enhancing Software Development Workflows

AWS announced a strategic agreement with Poolside, making its generative AI Assistant and foundation models available in Amazon Bedrock. This collaboration aims to improve software engineering tasks such as code generation, testing, and documentation.

Why it’s exciting: We believe there are opportunities to integrate Poolside with the high code functionality we are bringing to DPC, enhancing productivity and accelerating development cycles.

The Bigger Picture: Building a Data-First, AI-Driven Future

What we saw at AWS re:Invent 2024 reinforces a central theme: generative AI is not just a technology, but a force multiplier for data teams. For Matillion customers, these advancements are more than exciting—they are transformative. By pairing Matillion’s Data Productivity Cloud with AWS Bedrock’s innovations, teams can unlock:

• Faster time-to-value.
• Smarter insights from unstructured data.
• Scalable, cost-efficient AI workflows.

As generative AI continues to evolve, the collaboration between AWS and Matillion promises to lead the way in making data integration more intelligent, intuitive, and impactful than ever before.

Are you ready to transform your data workflows? Explore how Matillion and AWS Bedrock can drive your generative AI journey today.