2026 Modular Seoul Developer Meetup
Modular, a U.S. based AI unicorn, hosted its first Korea meetup with SqueezeBits. Read our recap of the first Seoul developer meetup!
Jun 19, 2026
Contents
The Meetup Opened With a Message From Modular CEO Chris Lattner 🙌Introducing Modular's Next-Generation AI Solutions, Mojo and MAX 👀SqueezeBits Case Study: Building a Real Service Pipeline With the MAX FrameworkNetworking Over Great Food and Drinks 🥂GPU Lucky Draw Event 🍀A Successful First Seoul Meetup, With More Opportunities Ahead 🔥Hello 😊
I'm Goeun Kang, Marketing Manager at SqueezeBits.
May is always packed with holidays like Children's Day and Parents' Day, and the nice weather brings even more events. SqueezeBits had one of its busiest months yet this May. We see it as a sign that more people are taking an interest in SqueezeBits, and we have been moving with a lot of energy and gratitude. 🏃♀️
On the evening of Tuesday, May 19, we hosted a slightly more special event: the first Korea developer meetup with Modular, a U.S.-based AI unicorn and strategic partner of SqueezeBits. Modular has recently been valued at over $1.6B and has raised more than $380M in funding. It has drawn strong global attention, but it is still relatively less well-known in Korea. So far, it seems most familiar to people with a background in compilers.
Modular is one of the companies that often comes up in conversations about U.S. AI startups. Since the rise of generative AI, developer communities have become increasingly interested in running AI faster and more efficiently. Modular has quickly built a presence in that conversation. You can think of it as an AI infrastructure company that helps AI run faster and more efficiently on any hardware.
This meetup was designed not only to introduce Modular but also to showcase a successful collaboration with SqueezeBits and to explore directions Korean companies are likely to be interested in.
The Meetup Opened With a Message From Modular CEO Chris Lattner 🙌
Modular CEO Chris Lattner sent a welcome video for everyone joining the Seoul meetup. Chris Lattner is a name many developers have heard at least once. He is best known as the creator of the Clang compiler, LLVM, and Swift, Apple's programming language. He has also had a major influence on next-generation compiler and optimization technologies that help models run efficiently across diverse AI hardware.
As this was Modular’s first offline meetup in Korea, Chris also shared his excitement. He said he was impressed by how quickly Korean developers are advancing, and he opened the meetup by expressing his hope that Modular and SqueezeBits could offer meaningful support to developers in Korea.
Introducing Modular's Next-Generation AI Solutions, Mojo and MAX 👀
Judy Heflin from Modular briefly introduced the company at a previous Efficient AI offline meetup. This time, she shared a deeper look at Modular, focusing on Mojo and MAX.
Big tech companies worldwide are releasing AI-specific chips one after another. Google has TPUs, AWS has Inferentia and Trainium, and NVIDIA has Blackwell. As each company develops its own AI hardware, AI computing environments are becoming more diverse. The challenge is that its diversity also makes development more complex. Code built for one chip often can’t be reused easily in another environment.
Modular's answer to this problem is Mojo.
Mojo is a new language that offers familiar Python-like syntax while delivering the high performance of a systems programming language. It aims to combine developer productivity with runtime performance and create an AI development environment that is not tied to specific hardware. Since its launch in 2023, Mojo has grown quickly, drawing interest from developer communities around the world and attracting more than 150,000 users.
After Mojo, Judy introduced MAX, an AI framework that helps teams run and operate AI models more easily in real service environments. MAX supports the full process from model development to optimization, deployment, and serving on a single platform. It is designed to deliver high performance across diverse hardware environments.
One especially notable announcement from the session was that Modular has open-sourced MAX's core technology. This creates a foundation for more developers to validate the technology and contribute improvements. It should also help expand the broader AI ecosystem.
The session also covered real performance examples. In an NVIDIA B200 environment, MAX recorded higher throughput than the existing vLLM setup. It also showed improved performance over a PyTorch environment on FLUX.1 and FLUX.2, both image generation models.
In short, Mojo and MAX make up Modular’s integrated ecosystem for an era of increasingly diverse AI hardware. Together, they let developers build and run AI systems without being tied to a single platform. The presentation’s key message was clear: reduce complexity while achieving both high performance and cost efficiency.
SqueezeBits Case Study: Building a Real Service Pipeline With the MAX Framework
Next, SqueezeBits CTO Taesu Kim shared how SqueezeBits improved generative AI performance with MAX, based on a real project carried out with Modular.
After forming a partnership with Modular, SqueezeBits has worked on a project to expand MAX beyond its existing use around large language models (LLMs) into image and video generation.
One of the most important factors in operating a generative AI service is reducing the time users spend waiting for results. But as AI models become more complex and hardware environments become more diverse, code that works well in one environment can often see major performance drops or unexpected issues in another.
To solve this, SqueezeBits optimized the execution structure of an image generation model for the MAX environment. Our team improved the overall processing flow by reducing repetitive operations and efficiently reusing previously computed results.
The session covered not only technical results, but also the practical strengths and limitations of MAX and Mojo that SqueezeBits experienced during the collaboration. It was a useful session because it gave attendees realistic insight into what it takes to apply new technology to a real service.
Networking Over Great Food and Drinks 🥂
Because the meetup took place in the evening, we also prepared networking time after the presentation sessions so attendees could talk more comfortably. A variety of food, desserts, beer, and wine helped create a relaxed atmosphere for conversation.
Attendees naturally discussed Mojo, MAX, AI optimization technologies, and other topics from the presentations while sharing their own experiences and insights. Compiler engineers and researchers who were especially interested in Modular also asked Judy questions directly and exchanged technical perspectives.
There was also strong interest in SqueezeBits' real project case using MAX. Attendees asked about the technical background and development process that could not fit into the presentation, and the discussions continued actively after the sessions ended.
The networking session lasted about an hour, but it went by faster than expected. Many attendees enjoyed the food together and kept conversations flowing naturally. It was a meaningful time for experts across different fields to connect through a shared interest in AI development and optimization.
GPU Lucky Draw Event 🍀
The meetup closed with a lucky draw event that many attendees had been looking forward to. We prepared an AMD RADEON RX 9070 and an NVIDIA RTX 5080 as prizes, which drew a lot of attention.
Winners were selected through a random draw based on the numbered tickets attendees received at check-in. These products are not easy to encounter at most events, so the atmosphere grew even more exciting during the draw. Since people from many areas of the hardware and software industries joined the event, everyone turned their attention to the stage when the prizes were announced. The two selected winners received their prizes on-site, and we also took commemorative photos with plenty of applause. 🙂 We hope the winners make great use of their GPUs! 😉
A Successful First Seoul Meetup, With More Opportunities Ahead 🔥
Because this was the first Modular Meetup held in Korea, we weren’t sure how much interest it would draw. But many more people joined than we expected, and thanks to them, we were able to wrap up the event successfully.
Conversations with attendees also reminded us why this meetup mattered. We met people who had installed and tried Mojo before the event, as well as compiler engineers who signed up as soon as they heard about the Modular Meetup.
We also heard comments like, “I wouldn’t have come across content like this without the meetup,” and “I’d definitely attend again if another Modular Meetup is held.” Many attendees also asked for deeper sessions on Mojo and MAX, underscoring just how strong the interest is in Korea’s developer community.
As a startup specializing in AI model compression and optimization, SqueezeBits continues to create activities that connect these deep-tech fields. We plan to keep creating spaces where people interested in AI infrastructure and optimization technology can learn and connect.
Thank you again to everyone who joined us for this Modular Meetup. We hope you will continue following the community activities SqueezeBits is building.😉 We share SqueezeBits news first on LinkedIn, so follow us there for updates!
Share article