A Review of the AWS Summit Seoul 2025
From May 14 to 15, 2025, I attended the AWS Summit Seoul, held at COEX. Although the event is held annually, this was my first time attending. As I mentioned in a previous post, I try to participate in as many conferences and workshops as possible to gain diverse experiences rather than settling into the comfort of the internal company environment.
Why I Attended
Currently, I’m working on generative AI services at my company. I’ve started serving quantized LLMs in an isolated (air-gapped) server environment, and I’m preparing to launch a POC-type LLM service utilizing our manufacturing database through LangChain. Before doing so, I was curious to see how AWS approaches generative AI and wanted to compare it with the service I’ve been building—thus, I decided to attend this conference.
Conference Experience
The event took place on the 1st and 3rd floors of COEX, and admission was free with prior registration, which I completed ahead of time. (Even though I registered a month in advance, the QR code didn’t arrive until just two days before the event, which made me slightly anxious.) Instead of just browsing the exhibition booths—which I felt wouldn’t provide deep insights—I focused my schedule around workshops. However, since the workshops ran from 11 a.m. to 6 p.m., I missed most of the main stage sessions after the keynote. It was a bit disappointing to head straight home after the workshop, so next time, I’m considering attending for both days to explore more leisurely.
As expected, the AWS ecosystem was vast. A wide range of topics was covered across various sessions. Still, I was particularly excited to attend the Generative AI Development Workshop using Amazon Bedrock, which made the keynote all the more enjoyable.
Workshop Field Notes
The workshop was held in Conference Room 402 on the 4th floor of COEX.
The main topics included how to write AI-powered reports using generative BI (QuickSight Q) and how to build and deploy an LLM assistant using AI Agents. At my company, we use Spotfire as our BI tool, while AWS offers QuickSight Q as a comparable solution. In practice, it felt somewhat lacking compared to Spotfire—possibly because I’m more familiar with the latter.
Still, AWS’s approach to automating data analysis and report generation using generative AI was noteworthy and promising.
The part that intrigued me the most was definitely the LLM service based on AI Agents.
Midway through the workshop, I had about 40 minutes of free time, so I visited the main stage on the first floor. I was lucky enough to find one empty seat and was able to join the session. Out of nearly 150 seats, every single one was taken—talk about perfect timing.
LLM Service Based on AI Agents
The official title of the session I attended was “Bedrock Multi-Agent Workshop in AWS Summit 2025 Seoul.” I first learned about Amazon Bedrock during the AW2025 event in March, where Hyundai AutoEver introduced their SD Brain project. Through a post-presentation conversation with one of the developers, I learned that the system had been built in the Bedrock environment in collaboration with AWS. That may have been what ultimately led me to attend this workshop.
The workshop did a great job of clearly explaining the concept of AI Agents—especially why we need to scale from Single-Agent to Multi-Agent setups, and how agents can collaborate with one another. Real examples made these concepts easy to understand.
This workshop also marked my first hands-on experience with Amazon SageMaker, which was quite eye-opening. Following the manual took some time, but seeing the process actually result in a deployed service was impressive. (Of course, as the saying goes, “the devil is in the details”—when building a real-world service, there would be much more to consider from the design stage onward.) Once again, I realized how far ahead the world outside of my company really is.By the time I finished going through the tutorial and completing the hands-on exercises, four hours had flown by, and I was so immersed that I didn’t even manage to take a single picture.
Final Thoughts
Attending AWS Summit Seoul 2025 gave me a clearer understanding of the direction generative AI is heading and the technological trends that are shaping it.
In particular, the Multi-Agent implementation using Amazon Bedrock provided valuable insights that will help shape the direction of my future work in manufacturing AI.
The AI Agent development project I plan to begin in the second half of this year will no doubt benefit from the experiences I gained through this workshop.