AWS re:Invent 2025 Recap - Vietnam Edition

Summary Report: “AWS re:Invent 2025 Recap - Vietnam Edition”

Event Objectives

  • Review the most important announcements and innovations introduced at AWS re:Invent 2025.
  • Help the Vietnamese developer community quickly understand new AWS services and architectural trends.
  • Provide insights into modern cloud technologies including Generative AI, Agentic AI, and next-generation data infrastructure.

Key Highlights

Generative AI and Agentic Systems on AWS

One of the main focuses of the event was the rapid development of Generative AI and autonomous AI Agents within the AWS ecosystem.
Participants were introduced to Amazon Bedrock and the Nova model family, which provide managed foundation models that developers can easily integrate into applications.

The session also highlighted Bedrock Agents, a framework designed to build intelligent systems capable of executing complex workflows autonomously. These agents support several important capabilities:

  • Orchestration and Flow management for coordinating multi-step tasks.
  • Memory systems that allow agents to retain contextual information.
  • Policies and Guardrails to ensure safe and controlled AI behavior.
  • Evaluation tools for monitoring and improving agent performance.

These components enable developers to build advanced AI-driven applications that can interact with external tools and data sources.

SageMaker Unified Studio and S3 Innovations

Another important topic was the evolution of the machine learning development environment on AWS.
The newly introduced SageMaker Unified Studio provides a centralized workspace where Data Engineers, Data Scientists, and AI Engineers can collaborate within a single integrated IDE.

In addition, several enhancements to Amazon S3 were presented:

  • S3 Tables enable native storage using the Apache Iceberg table format, simplifying large-scale analytics workflows.
  • S3 Vector introduces built-in vector storage capabilities, which significantly reduce the cost of managing embeddings compared to traditional vector databases.

These updates demonstrate AWS’s effort to unify data storage, analytics, and AI development in a single platform.

The event also introduced new capabilities of OpenSearch Serverless, particularly its integration with modern AI systems.

Key concepts included:

  • Integration with Model Context Protocol (MCP).
  • The concept of Agentic Memory, allowing AI systems to retain and utilize search knowledge.
  • Deployment of specialized agents designed for data analysis tasks.

A live demonstration showed how a Flow Agent could automatically analyze sales data and extract insights by interacting with the search system and analytical tools.

Advanced RAG and Multimodal AI

Beyond traditional text-based AI applications, the event explored the evolution of Retrieval-Augmented Generation (RAG) toward multimodal capabilities.

Important technologies discussed included:

  • Nova Multimodal Embeddings, which convert images and videos into vector representations.
  • Bedrock Data Automation, enabling automatic extraction and structuring of information from multimedia sources.

These innovations allow AI systems to work with a much broader range of data types such as images, videos, and audio, expanding the possibilities for intelligent applications.

AI Infrastructure and SageMaker Deep Dive

The final session focused on the infrastructure required to train and deploy large-scale AI systems.

Several tools within the SageMaker ecosystem were introduced:

  • SageMaker HyperPod, designed for managing large GPU clusters used in deep learning workloads.
  • SageMaker MLflow, supporting experiment tracking and lifecycle management of machine learning models.
  • Bi-directional streaming capabilities, which enable real-time Voice-to-Voice AI applications.

These features highlight AWS’s focus on providing scalable infrastructure for the next generation of AI systems.

Some event photos

Event Check-in Networking with attendees

Overall, the event provided a comprehensive overview of the major announcements from AWS re:Invent 2025, particularly in the areas of Generative AI, agent-based architectures, and modern data platforms. The sessions also offered practical insights into AWS services that could be applied to real-world systems such as the NutriTrack architecture, while giving participants the opportunity to connect directly with AWS Solution Architects and members of the local cloud community.