Back to all Re:Invent 2024
Table of Contents
AI & ML
 |
This talk explores the importance of building explainable AI models using Amazon SageMaker, highlighting the need for transparency in AI decision-making and the various techniques available for global and local model explainability. The speaker discusses how Amazon SageMaker simplifies the implementation of explainable AI, providing tools for pre-deployment bias detection and post-deployment explainability monitoring to ensure ethical and responsible AI practices.
|
 |
This talk explores how to build large language models for low-resource languages, such as Cantonese, using AWS SageMaker HyperPod clusters. The speaker discusses the challenges of training models for these languages, the data preparation process, and the architecture and performance of the resulting Cantonese language model, which is open-sourced and available for further development.
|
Analytics
 |
This talk explores the role of analytics engineering in bridging the gap between data producers and consumers, and how analytics engineers can leverage AWS services to solve real-world problems in a financial company. The speaker highlights the key skills, tools, and future prospects of the analytics engineering career, emphasizing its importance in the emerging AI era.
|
Compute
 |
The presenters discuss the various container services offered by AWS, including EC2, ECS, EKS, Fargate, and App Runner, and how to choose the right service based on factors like cost, complexity, and flexibility. They also provide a detailed comparison of the different services and their trade-offs, as well as a demo and resources for further exploration.
|
 |
The presentation explores the challenges of troubleshooting a production issue in an Amazon EKS (Elastic Kubernetes Service) environment, where a seemingly simple problem turned into a months-long journey of investigation, root cause analysis, and eventual resolution. The speaker shares valuable lessons learned, emphasizing the importance of considering all components, including external tools and agents, validating compatibility, and maintaining patience and persistence in the face of complex and elusive issues.
|
 |
The speaker discusses Honeycomb's migration to AWS Graviton, detailing the performance and cost benefits, as well as the challenges and strategies involved in transitioning their infrastructure to the new ARM-based architecture. The presentation covers Honeycomb's approach to testing, rollout, and ongoing management of the Graviton platform across their various services and workloads.
|
 |
This presentation explores the use of Amazon Q Developer, a generative AI solution, to manage AWS costs effectively. The speaker demonstrates how Amazon Q Developer can provide quick and accurate answers to cost-related questions, while also highlighting the continued need for human cost expertise in complex scenarios.
|
 |
In this talk, the speaker discusses how a data engineering team at an ad media company was able to significantly speed up their ETL processing by rethinking their data pipeline, leveraging Apache Spark on Amazon Athena, and addressing key performance bottlenecks. The talk provides practical insights and optimization strategies that can be applied to improve the performance of data pipelines in various contexts.
|
 |
The presentation discusses scaling machine learning with containers on AWS, covering lessons learned from the speaker's experience at Instrumental, a manufacturing optimization platform. It highlights how container-based solutions, such as SageMaker and Lambda, can address challenges in research speed, productization, and scalability, while also providing insights on cost optimization and observability.
|
Developer Experience
 |
The presenter discusses the use of Amazon Q, an AI-powered programming assistant, to boost developer productivity. She covers deployment steps, key use cases, lessons learned, challenges and solutions, and her personal experience with the tool's code transformation capabilities.
|
 |
This presentation showcases Epimetheus, a new open-source project that utilizes Amazon Bedrock and natural language processing to simplify the creation of PromQL queries on Grafana and Prometheus. The tool aims to address the challenge of remembering complex PromQL syntax, especially in emergency situations, by allowing users to input natural language queries that are automatically transformed into the corresponding PromQL.
|
 |
The session explores the challenges of data collaboration and introduces AWS Clean Rooms, a secure platform that enables organizations to analyze data insights without sharing raw data. The presentation demonstrates the setup and usage of AWS Clean Rooms, highlighting its features, analysis rules, and the ability to collaborate on data-driven decisions while maintaining data privacy and control.
|
 |
The presenters discuss techniques for transforming large language models into domain-specific experts, including continued pre-training, fine-tuning, prompt engineering, retrieval-augmented generation (RAG), and the use of agents. They provide practical examples and guidance on how to leverage these techniques using AWS services like Amazon Bedrock to build customized AI solutions for various business needs.
|
 |
The presenters showcase the latest version of AWS Amplify, a framework that enables developers to build cross-platform applications with ease. They demonstrate how to leverage Amplify's features, such as authentication, data management, and AI-powered functionality, to create a travel planning application with a seamless user experience.
|
 |
This session discusses how the AWS community in Brazil used AWS Amplify to build a custom event management platform for organizing community events and managing speaker submissions. The speakers highlight the challenges faced by community organizers, the benefits of using Amplify for rapid development and deployment, and showcase other community-built tools that leverage AWS services to support event organization and networking.
|
 |
This talk explores how the Pacific Northwest National Laboratory leveraged Amazon Bedrock and large language models to analyze over 4,000 Jira tickets and uncover valuable insights, such as emerging trends and common customer issues. The speaker shares their journey, including the challenges of data preparation, knowledge base creation, and prompt engineering, and highlights the potential benefits of this approach for improving customer support and service delivery.
|
 |
This presentation discusses the strategy and technology behind platform engineering, a critical approach for managing complex cloud infrastructure and accelerating development. The speaker shares Fujitsu's experience in implementing a platform engineering solution, highlighting the benefits, challenges, and future roadmap for this transformative approach.
|
 |
The video discusses how AWS re:Invent 2024 will focus on boosting productivity with Amazon Q Developer agents, including features like code understanding, visualization, code transformation, and code generation. The presenter showcases various capabilities of Q Developer and emphasizes the importance of prompt engineering and using prompt templates to ensure consistency in the generated code.
|
 |
This session explores the use of AWS Bedrock, a fully managed serverless service for building and scaling generative AI applications, along with Amazon Q, a coding assistant, to create an agentic workflow that can help an AWS Solutions Architect with tasks like querying the AWS Well-Architected Framework, generating code, and creating architectural diagrams. The presenters demonstrate how these tools can be integrated into a developer's workflow to enhance productivity and empower individuals to build innovative solutions.
|
 |
This presentation covers modern CI/CD strategies and tools, emphasizing the importance of automation, code signing, deployment consistency, and the use of continuous configuration and feature flags to decouple deployment from release. The speakers discuss techniques to manage infrastructure drift, enforce organizational standards, and leverage generative AI to empower developers in the deployment decision-making process.
|
 |
This presentation explores the risks and real-world learnings of using Infrastructure as Code (IaC) and AI, particularly Generative AI, from the perspectives of three personas: DevOps/SRE, Security Engineer, and Developer. The key takeaways emphasize the importance of context, security, collaboration, and leveraging AI tools with caution to avoid potential pitfalls.
|
 |
This talk explores how Amazon API Gateway can enhance application responses by leveraging features like caching and throttling. The speaker demonstrates a practical implementation using AWS CDK, highlighting the benefits of these techniques in improving user experience and maintaining application performance.
|
 |
This talk discusses the speaker's experience in leading the migration of an airline notification service from a mainframe to the AWS cloud. The speaker highlights how they leveraged Amazon's leadership principles, community support, and a focus on delivering practical solutions to overcome various challenges and successfully complete the project.
|
 |
This session showcases a comprehensive AI-powered travel assistant solution that leverages PostgreSQL with vector support, AWS Bedrock, and seamless integration with messaging platforms like WhatsApp. The presented approach demonstrates how to build a scalable, personalized, and multilingual customer support system that combines knowledge retrieval, data-driven actions, and prompt engineering techniques to deliver an enhanced travel experience.
|
 |
This talk provides a comprehensive overview of the evolution of infrastructure as code tools, including AWS CloudFormation, Terraform, and AWS CDK. The speaker delves into the strengths and tradeoffs of each tool, offering practical guidance on how to choose the right one based on the specific needs of your organization and project.
|
 |
The video discusses the benefits of application modernization, the challenges of dealing with legacy code, and how Amazon Q Developer can help automate and accelerate the code migration process. The speakers also provide a structured approach to modernization using the ModAx methodology and highlight various AWS modernization pathways and refactoring strategies to help organizations transform their applications.
|
 |
This session explores how to build scalable and engaging community applications using Amazon Q, a generative AI service from AWS. The speakers share their journey of leveraging various AWS services and techniques to create a comprehensive event planning solution that combines data from multiple sources, identifies optimal event locations and speakers, and generates promotional content.
|
 |
The session introduces GenASL, a generative AI-powered application that enables multimodal American Sign Language (ASL) avatars with bidirectional translation capabilities. The solution leverages foundational models and multimodal AI techniques to provide seamless communication between Deaf/hard-of-hearing users and hearing users, addressing the global shortage of sign language interpreters.
|
 |
Kristi Perreault and Matt Coulter, co-founders of Teach Me AWS, share their journey in building an educational platform on AWS to address the challenges faced by developers in learning AWS services and their real-world use cases. They discuss their architectural decisions, the use of serverless technologies, and their plans to continue growing the platform while maintaining a cost-effective and flexible approach.
|
 |
This talk explores how generative AI can supercharge DevOps practices by enhancing problem-solving, increasing operational efficiency, enabling adaptive learning, and providing unprecedented scalability. The presenters demonstrate how tools like Amazon Q Developer and Amazon Bedrock can accelerate engineering productivity, address software delivery bottlenecks, and streamline incident response, ultimately transforming the DevOps lifecycle.
|
Security
 |
This session explores the concept of security invariants, which are preventative controls that are strictly enforced to reduce the burden on development and operations teams. The speakers discuss various tools and techniques for implementing security invariants, such as service control policies, resource control policies, and declarative policies, and provide practical examples of how to define and deploy these controls.
|
 |
This talk presents a system called 'Liver' developed by Kaminashi, a Japanese company, to enable developers to self-remediate AWS Security Hub findings. The system aggregates findings, provides clear explanations and remediation steps, and assigns responsibility to individual developers, fostering a security-conscious culture and reducing the burden on the security team.
|
 |
This talk explores the security challenges of Amazon S3, covering Red Team exploits and Blue Team defenses. The speaker discusses various encryption options, privilege escalation tactics, and proactive measures to secure S3 buckets, including AWS Config rules, EventBridge automation, and VPC endpoint policies.
|
 |
This talk covers how an enterprise organization, Simply Business, secured 50 million requests per month using AWS-based authorization with Amazon Verified Permissions and Cedar. It discusses the implementation of external role-based access control, internal attribute-based access control, and internal role-based access control, highlighting the challenges and lessons learned along the way.
|
Serverless
 |
The talk explores the challenges and best practices of building scalable serverless architectures, focusing on a case study of a live scoring system for a global sporting event. The speaker shares key learnings on identifying and addressing scalability boundaries, leveraging patterns and blueprints, and the importance of performance testing and flow control to unleash the full potential of serverless scalability.
|
 |
The presentation explores the transition from a single-tenant to a multi-tenant serverless application architecture, highlighting the trade-offs and challenges faced by the speaker's team at PostNL. The key focus is on maintaining mission-critical service delivery while optimizing deployment time, cost, and resource utilization through a gradual migration approach.
|
 |
The video presents an AI-powered serverless solution for a garment printing company to efficiently search and retrieve design images based on various attributes, leveraging the power of generative AI models and serverless cloud infrastructure to deliver a cost-effective and scalable solution.
|
 |
This video discusses how to use AWS Powertools for Lambda to enhance observability, architectural patterns, and resiliency in serverless applications. The speaker demonstrates how Powertools can help with metrics, tracing, caching, idempotency, and structured logging, providing a comprehensive overview of the library's capabilities.
|
 |
This talk presents a case study of how a team at Chemist4U, an online pharmacy in the UK, built a serverless solution to integrate with a new clinic partner in just 8 weeks. The key factors discussed include leveraging serverless technologies, following a lean and agile development process, and focusing on the core business logic to rapidly deliver a minimum viable product.
|