Event 1

Summary Report: “AWS Cloud Mastery Series #1 - AI/ML/GenAI on AWS”

Event Objectives

  • Provide an overview of AI/ML/GenAI capabilities on AWS and how they can be applied in real-world use cases.

Speakers

  • Lam Tuan Kiet – Sr DevOps Engineer, FPT Software
  • Danh Hoang Hieu Nghi – AI Engineer, Renova Cloud
  • Dinh Le Hoang Anh – Cloud Engineer Trainee, First Cloud AI Journey
  • Van Hoang Kha – Community Builder

Key Highlights

1. Generative AI on Amazon Bedrock

Foundation Models (FMs):
AWS supplies a library of fully managed FMs from multiple leading AI providers (Anthropic, OpenAI, Meta, etc.), allowing users to adapt them to various tasks without building models from scratch.

Prompt Engineering:
Understanding how to guide models effectively through different prompting strategies:

  • Zero-shot: Model receives only the task description.
  • Few-shot: Model is given a handful of examples to mimic.
  • Chain-of-Thought: Encouraging the model to reveal reasoning steps for more accurate outcomes.

Retrieval Augmented Generation (RAG):
Enhances model outputs by injecting external knowledge:

  • R – Retrieval: Pulls relevant data from a knowledge store.
  • A – Augmentation: Adds this data as context in the prompt.
  • G – Generation: Model creates a more grounded and accurate answer.
  • Use cases: Chatbots with knowledge bases, contextual search, and near-real-time summary generation.

Amazon Titan Embeddings:
Lightweight embedding model that turns text into dense vectors for similarity search and RAG workflows, with multilingual support.

AWS AI Services: Ready-made AI APIs for common tasks:

  • Rekognition – Image/Video analysis
  • Translate – Language translation
  • Textract – Extract text + layout from documents
  • Transcribe – Speech-to-text
  • Polly – Text-to-speech
  • Comprehend – NLP insights
  • Kendra – Intelligent enterprise search
  • Lookout – Anomaly detection
  • Personalize – Recommendation systems

Demo Highlight:
A simple face-recognition application (AMZPhoto) demonstrating how AI services integrate with user-facing apps.

2. Amazon Bedrock AgentCore – Building Production-Ready AI Agents

A new framework enabling teams to operate AI agents reliably at scale:

  • Execute and scale agent workflows securely
  • Manage long-term memory
  • Grant fine-grained identity & access control
  • Integrate with tools like Browser Tool, Code Interpreter, Memory Store
  • Provide observability and auditing
  • Support popular agent frameworks (CrewAI, LangGraph, LlamaIndex, OpenAI Agents SDK, etc.)

Key Takeaways

  • Bedrock as the central GenAI platform: Easy access to many FMs in one place.
  • Prompting + RAG as customization tools: Improve model relevance with context and examples.
  • Embeddings for smarter search: Titan Embeddings boosts retrieval accuracy for knowledge-driven applications.
  • Pretrained models accelerate development: No need to build everything manually.
  • AgentCore reduces deployment complexity: Handles scaling, memory, identity, and monitoring for agentic systems.

Applying to Work

  • The concepts from the session (especially RAG and AgentCore) can be directly aligned with future internal projects involving GenAI-powered features.
  • Understanding AWS AI services helps in selecting the right tools for different application needs, speeding up development cycles.
  • The knowledge gained about prompt engineering can be applied to improve the performance of AI models in various scenarios.

Event Experience

  • Placed top 5 in the end of event Kahoot Quiz and got a picture with the speakers
  • Formed a small collaborative group called “Mèo Cam Đeo Khăn”, combining members from “The Ballers” and “Vinhomies”.

Some event photos

Event Photo 1 Event Photo 2 Event Photo 3