Event 1
AWS re:Invent Recap HCMC 2026
Event Objectives
- Tech Updates: Capturing the latest announcements from global AWS re:Invent, specifically the Nova model series, Amazon Bedrock, and MLOps infrastructure.
- Solution Optimization: Exploring the implementation of Vector Databases and Multimodal Retrieval to solve complex data challenges.
- Local Engagement: Direct networking with AWS experts regarding application modernization and the new Vietnam DirectConnect POP infrastructure.
Speakers
- Prapti Gupta - AWS Speaker & Expert (Led the session on Vector Database on S3).
- Solutions Architects from AWS Vietnam.
Key Highlights
- Keynotes on ML/AI & Innovation: AWS’s strategic vision for driving innovation through the power of data and AI.
- Vector Database on Amazon S3: Method for storing and managing data vectors directly on S3, optimizing cost and scale for AI applications.
- Natural Query Language in OpenSearch: Using natural language to query data, bridging the gap between users and complex data systems.
Generative AI Ecosystem and MLOps Infrastructure
- Upgrading GenAI Applications: Leveraging the latest features of the Nova model family and Amazon Bedrock to create smarter applications.
- Resource Management via MCP: Mechanism for efficient interaction and coordination of AWS resources through the Model Control Plane (MCP).
- Multimodal Retrieval: A breakthrough in Bedrock Knowledge Bases, allowing search and synthesis across various data types (images, text, tables).
- Large-scale MLOps Infrastructure: Professional workflows for training and deploying machine learning models on the Amazon SageMaker AI platform.
Key Takeaways
Professional Knowledge
- Architecture Flow: Understood the processing flow from storage (S3) and querying (OpenSearch) to training and execution (SageMaker/Bedrock).
- Power of Multimodal: How to combine information from multiple data formats to enhance AI accuracy.
Systems Thinking
- Realized that AI technology is not just about models, but also the entire ecosystem of infrastructure and governance (MLOps) to ensure scalability.
Applying to Work
- SmartInvoice Shield Project:
- Apply Multimodal Retrieval to handle complex input invoices (blurry photos, structured PDFs).
- Replace or supplement relational databases with Vector Databases to accelerate cross-referencing and intelligent search.
- Build standard MLOps workflows to manage OCR model versions more effectively.
Event Experience
Attending the event at the Bitexco office was a professional and inspiring experience:
- Directly listened to the passionate sharing by speakers about optimal storage solutions.
- Valuable networking opportunities with the Cloud & AI engineering community in Vietnam.
Lessons Learned
- Understanding the architecture flow is more important than just knowing how to use individual services.
- Cost optimization from the design phase (such as using S3 for Vector DB) is a survival factor for large-scale Cloud projects.
- Networking is not just for connecting, but also for validating one’s own technical assumptions through the experience of others.
Some event photos

Tech Solution Updates

AWS Community Memories 1

AWS Community Memories 2
Overall, the re:Invent Recap 2026 event provided me with a clear “roadmap” to shape the AI direction for the SmartInvoice Shield project and my upcoming career path.