/Events /AgentCon 2025 - Ho Chi Minh /Schedule
AgentCon 2025 - Ho Chi Minh
Thursday, September 11
Registration
Plenary ·Welcome to AgentCon Ho Chi Minh by Dux Raymond Sy
Plenary ·Human in the Loop: Guardrails for AI’s Wild Side by Dona Sakar
2nd Floor | Lecture Hall A5 ·Bringing AI Agents to Life with Azure AI Foundry
AI agents are moving beyond text boxes into real-world interactions — combining models, tools, and avatars to deliver human-like experiences. In this session, we’ll explore how to build an end-to-end solution with Azure AI Foundry, weaving together model selection, instruction design, and tool integration into a single workflow. As a live scenario, we’ll create an AI agent that can order a pizza through a connected MCP server — and then bring it to life with a real avatar for natural, conversational interactions. You’ll leave with a clear view of how to turn prompts and models into full-fledged, real-world AI agents.
Henk is a Cloud Advocate specializing in Artificial intelligence and Azure with a background in application development. He is currently part of the AI cloud advocate team and based in the Netherlands. Before joining Microsoft, he was a Microsoft AI MVP and worked as a software developer and architect building lots of AI powered platforms on Azure.
AI Unleashed: The Foundry and studio shaping tomorrow's creativity
The Foundry and other AI studios are making it easier for businesses to be creative and innovative. For example, companies can use these platforms to design virtual try-on tools for clothing, allowing customers to "wear" items online before buying. They can also develop chatbots for customer support, helping to answer questions quickly and efficiently. These tools make processes smoother, save time, and create better experiences for customers.
From Singapore. I work with SharePoint and Power Platform. I'm a Microsoft MVP in Singapore who plans events and is involved in community things. Check me out on YouTube at https://www.youtube.com/@msisgreat. I'm one of the organizers of the Microsoft Meetup community with over 7500 members islandwide (http://www.meetup.com/mssgug/).
Agentic Search with small models - Jan-nano, Lucy
Introducing Jan-nano, Lucy two latest work of Menlo Research lab.
Alan Dao is an AI researcher and builder passionate about creating systems that push the boundary of intelligence and usability. His work spans multimodal models, robotics, and agentic workflows—focusing on real-time reasoning, spatial understanding, and open-source accessibility. He is the creator and co-author of several impactful systems and research works, including: Jan-nano, a 4B tool-using language model designed for search and question-answering agents Ichigo, a realtime voice assistant using early fusion multimodal modeling AlphaMaze and AlphaSpace, frameworks that bring spatial and symbolic reasoning to LLMs for visual tasks and robotic manipulation ........ His research has been published in major forums like arXiv and Interspeech 2025, where his recent work Speechless explores instruction-tuned models for low-resource languages—without requiring any speech. Alan’s open-source contributions have helped shape the evolving agentic ecosystem, with systems like Nitro (now Cortex) powering efficient LLM inference at scale. His projects are available on GitHub, and his thoughts are shared on Twitter, Threads, and his blog alandao.net.
Build AI voice agents to automate your workflows
Build AI voice agents to automate your workflows
In this session, you’ll learn how to build your own voice-controlled AI assistant—just like Jarvis from Iron Man—that can listen, understand, respond in real time, and more importantly, take intelligent actions directly on the web using only your voice.
With the power of OpenAI Realtime API and the Browser Use feature, you’ll explore how to:
Convert your speech to text instantly
Understand your intent using AI reasoning
Respond in a natural voice
Interact with websites: search, navigate, click, fill in forms, and more
Imagine giving voice commands like:
“Post this content to all my social media accounts.”
“Search Bing for the latest AI trends.”
“Read my messages and summarize the key points.”
“Find a nice restaurant near me with good reviews.”
Whether you’re a developer, a tech enthusiast, or a productivity hacker, this session will help you build your own intelligent voice agent to automate everyday tasks in a truly hands-free way
Vu Tong has nearly 2 years of experience in AI, focusing on LLMs. He’s also been active in Vietnam’s tech communities for 4 years, including Google Developer Groups HCMC, Women Techmakers HCMC, Google Developer Student Clubs, and UNDP YDCC,...
Agentic Software Development - Some Dos and Donts
In this session, I will demo live step by step with coming up a micro-SaaS idea, honing the idea with ChatGPT to derive the software spec and pass it down to Claude Code to get the multiple agents build a software from end to end.
I will also share my learning notes with agentic software development and vibe coding.
I am a software engineer turned into community builder and educator. I am passionate with spreading knowledge about tech to the community and nurturing the next gen of young tech talent.
Code Guardian: Building Multi-Agent AI Systems for Autonomous Security Operations
Attackers exploit vulnerabilities in days, but organizations take months to patch them, creating a 97-day average exposure window. With over 29,000 new CVEs reported in 2023, human-scale security is no longer a viable defense against machine-scale threats. This talk demonstrates a practical solution: a multi-agent AI system that automates the security lifecycle. This system autonomously detects vulnerabilities with fine-tuned LLMs, generates patches, and validates fixes. We will walk through the architecture of this "Code Guardian" and learn its capabilities.
The key takeaways from this talk are:
- Learn practical techniques to fine-tune code analysis models for the security system lifecycle.
- Discover how to architect a reliable multi-agent system with robust error handling, agent coordination, and fail-safe mechanisms for production environments.
- Implement strategies for managing edge cases where automated patching fails, ensuring the system remains both effective and safe.
Rashmi is a Machine Learning Engineer at Patchstack and a researcher at MIT CSAIL, working in the intersection of cybersecurity and artificial intelligence. She has six years of industrial experience, having brought ideas to life at pre-seed startups and contributed to impactful redesigns and features at established industry giants. Beyond coding, Rashmi finds inspiration in capturing the wonders of the cosmos through her telescope and engaging in board games with friends.
Lunch
2nd Floor | Lecture Hall A5 ·Chat with your Data Using Vector Search
Discover the Power of Vector Search with Azure SQL! There's a new way to search — and it's smarter than ever. Join us for an exciting session where we’ll explore how Azure AI Services and LangChain seamlessly integrate with Azure SQL’s vector search capabilities. Learn how to build chat-enabled applications that interact directly with your operational data, unlocking new possibilities for intelligent, real-time insights. Whether you're modernizing existing systems or building from scratch, this session will show you how to bring conversational AI to your data — effortlessly.
Pio Balistoy is a Filipino Microsoft MVP for Data Platform based in Singapore. Currently a Lead Data Platform Consultant at Pythian, he helps businesses drive transformation through data, analytics, and cloud. He frequently organizes and speaks at user groups and conferences in the APAC region.
Building hyperautomation workflow with multi-agents
Discover how to orchestrate robust hyperautomation workflows using autonomous multi-agent systems. This session demystifies agent collaboration patterns, showcases architecture blueprints, and highlights integration best practices across enterprise use cases. Attendees will walk away with clear insights into deploying agent-powered automation—from triggering actions to feedback loops and optimization.
Jonathan Viet Pham is a builder, speaker, and founder driving innovation at the intersection of AI and enterprise automation. As CEO of Diaflow, he leads the creation of agent-powered solutions that scale operations, reduce costs, and transform workflows. With years of experience in designing intelligent systems, Viet blends engineering depth with business insight to push forward the frontier of hyperautomation.
From Zero to PizzaBot: Building AI Agents with Azure AI Foundry
In this hands-on workshop, you’ll learn how to build intelligent, domain-specific AI agents using Azure AI Foundry Agent Service. Starting from a simple “hello world” agent, you’ll progressively enhance it with system prompts, custom instructions, and external knowledge via RAG (Retrieval-Augmented Generation). You’ll then extend your agent with tool calling, enabling it to run custom functions like a pizza calculator, and finally integrate with external services through the Model Context Protocol (MCP) for live menu and order management. By the end of the workshop, you’ll have a fully functional Contoso PizzaBot—an AI assistant that can answer questions, recommend pizzas, manage orders, and stay grounded in real business data.
Henk is a Cloud Advocate specializing in Artificial intelligence and Azure with a background in application development. He is currently part of the AI cloud advocate team and based in the Netherlands. Before joining Microsoft, he was a Microsoft AI MVP and worked as a software developer and architect building lots of AI powered platforms on Azure.
Playing with Fire: Unethical AI
AI is powerful, accessible, and incredibly easy to misuse. In this talk, I’ll walk through a series of intentionally provocative projects I’ve built, like an app that writes breakup texts using GPT or one that helps people generate fake excuses to skip meetings. These aren’t just jokes, they’re experiments meant to expose how AI can quietly enable unethical behavior.
Through these weird and sometimes uncomfortable tools, I’ll explore the ethical grey zones we often ignore: how small design choices shape user intent, how automation detaches us from responsibility, and how quickly things can spiral when no one’s watching.
This session isn’t about solutions, it’s about asking better questions. If you’re building with AI, you need to know just how easy it is to cross the line. Sometimes without realizing it.
Elian Van Cutsem or "ElianCodes" is a full-stack developer, creative technologist, and all-around internet troublemaker based in Belgium. He’s the CTO at Vulpo and works in Developer Relations at React Bricks, focusing on developer experience, documentation, and education. Previously, he was a Software Engineer and core maintainer at Astro. Elian is also a co-organizer of several major JavaScript communities and events, including BeJS, React Paris, React Brussels, and React Africa. Alongside his professional work, he creates provocative and experimental AI projects, designed to challenge our assumptions about ethics, automation, and human behavior. His work blends code and commentary, often in ways that are both hilarious and uncomfortably real. Elian is known for his passion, puns, and offbeat sense of humor. His talks combine technical insight with a strong dose of creative chaos—guaranteed fun on stage.
MCP & A2A: Powering the Agentic AI Stack for Enterprise-Grade Apps
Model Context Protocol (MCP) and Agent-to-Agent (A2A) form the foundational layer of secure, scalable, and reliable agentic AI systems. MCP enables context-driven execution across distributed agents, while A2A ensures trusted identity, encrypted communication, and autonomous coordination. Together, they unlock a new class of enterprise applications—where intelligent agents safely orchestrate workflows, share intent, and adapt in real time.
Thang is a Technical Manager at NashTech Vietnam, a Microsoft Azure MVP for 6 years in a row. He has over 18+ years of experience in software consulting, design, and development in outsourcing, product, and startup companies during his career. He currently works at NashTech with software experts and consultants in the company to design, develop and deliver various types of enterprise software such as Environmental, Social, and Governance (ESG), real estate, social networks, media marketing, gym & healthcare, and a variety of other business industries. He is in charge of a group of technical architects who help his customers migrate and modernise their legacy systems using cutting-edge technologies such as cloud computing, cloud-native platform, serverless, WebAssembly/WASI, and Agentic AI. In sparse time, he usually learns and shares his knowledge via GitHub, Twitter, and Facebook. And he is a creator and maintainer of many projects and sample code on his GitHub, as well as the founder of the Microservices in Vietnam community.
Break
2nd Floor | Lecture Hall A5 ·Text-to-SQL AI Agent with AWS Bedrock and LangGraph
Getting data just by asking questions in plain English? That's the dream of any business user in a tech company. But taking an exciting idea like that and turning it into a reliable, production-ready text-to-SQL agent - comes with its own set of fascinating headaches. At Parcel Perform, we faced some production and scaling challenges as we built our text-to-SQL solution. We'll dive into how we tackled issues like ensuring data accuracy with ambiguous inputs, optimizing for performance to prevent costly and slow queries, and effectively managing the operational costs of our AI initiatives. Join us to learn from our experience in making text-to-SQL work reliably at scale.
Vy is the AI Lead at Parcel Perform, where she leads the development of advanced AI solutions to solve complex business challenges. As an enthusiastic learner, she is dedicated to mentoring data professionals and building high-performing, collaborative AI teams. She enjoys turning ideas into practical solutions and believes in the power of continuous learning to drive meaningful progress.
Agentic AI: Aligning Intelligence and Ensuring Safety in Autonomous Systems
Agentic AI and some intro about AI Alignment, safety, and the potential threats of Agentic AI Misalignment
With 11+ years of experience in Natural Language Processing (NLP), Machine Learning (ML), Data Science, and Object-oriented programming, I bring a strong foundation to any project. My expertise extends to MLOps, and software development cycle with over 5 years of hands-on experience. As a part-time researcher at DataScience-Works lab in Australia, I stay at the forefront of industry trends and advancements. Now I'm the Head of AI at Obello, San Francisco & advisor to AI Safety Vietnam
It’s Not As Hard As You Think: Tales from the lone techie at a 2-person AI EdTech startup
What does it really take to build a real world agent-powered product from scratch? In this talk, Jeremiah shares his journey building Playbook, an AI EdTech startup, as the only engineer on a two-person team.
Along the way he rolled his own memory system, wrestled with LangChain vs. custom orchestration, migrated to Azure AI Foundry from "vanilla" Azure, and stood up agents to act as guides for students. He'll share lessons that go beyond frameworks: how to use AI (and common sense) to guide dev experience, human-agent interaction, and security all shaped what worked (and what didn’t).
You’ll leave with practical takeaways from real-world use — how to assemble with agent frameworks, when to prioritize safety and memory over features, and why trust is the hardest part of any AI system.
Jeremiah Marble is a technical AI product leader, author, and speaker who focuses on building high-quality, customer-centric software in the AI era. He’s currently writing code as the CTO of Playbook, an AI-based startup bootstrapping in the EdTech space. He also speaks, teaches, and leads workshops on AI, agents, and responsible AI (RAI) at 3rd Rodeo AI. Previously, Jeremiah held leadership roles at Mozilla and Microsoft—where he helped grow the Windows Insider Program from a whiteboard idea to more than 22 million people worldwide. An advocate for tech for good and social entrepreneurship, Jeremiah launched “Do The Thing Academy” and co-wrote Model 47: A Startup Storybook to help non-traditional founders turn napkin ideas into responsible businesses. He founded the ethical fashion line Towards Arcadia, empowering single mother tailors worldwide - including in Senegal, Mexico, and with Afghan refugees in Seattle. Previously, Jeremiah worked for the UN in Africa and Asia, was a Fulbright Scholar to Costa Rica, and volunteered in the Dominican Republic with the Peace Corps. He earned an MBA from Wharton, MA in International Studies from U Penn, and CS undergrad from Columbia.