How vLLM Built an AI Inference Community Across Asia in 2025
A recap of vLLM's four flagship 2025 community events across Singapore, Bangkok, and Malaysia — and how they built Asia's AI inference movement
If you had to pick one year that AI inference stopped being a niche engineering conversation and became a full-blown community movement in Asia, it would be 2025. Over the course of twelve months, the vLLM community — together with partners like AMD, Embedded LLM, SGInnovate, and Red Hat — ran four events across three countries, brought together open-source maintainers and enterprise leaders in the same room, and gave regional LLM builders a stage they'd never had before.
Here's a look back at the journey.
vLLM Asia Developer Day: Where It All Began
Singapore · 3 April 2025
Every community has a founding moment, and for vLLM in Asia, it was the Inaugural vLLM Asia Developer Day, co-organized with SGInnovate, AMD, and Embedded LLM. What started as a single-day gathering to talk about LLM inference turned into something none of us fully expected: a full house, then a waitlist, then a scramble for extra evening slots just to fit everyone in.
The numbers tell the story on their own. Registration interest pushed close to 400 sign-ups, and on the day itself, more than 300 people showed up — packing out the morning and afternoon sessions so quickly that SGInnovate had to reopen the listing just for evening slots.
What made the day special wasn't just the turnout — it was the lineup. The event opened with the vLLM Asia Community Launch, followed by a "State of vLLM" talk from core vLLM committers Chen Zhang and Cyrus Leung, and a deep dive into running vLLM in production. The afternoon shifted into hands-on territory with AMD ROCm optimization sessions and a "Zero to Production" workshop where attendees deployed models on live AMD GPU infrastructure.
But the moment that really defined the day was the roundtable that brought Singapore's own LLM builders onto one stage together for the first time — engineers from AI Singapore's SEA-LION, A*STAR's MERaLiON, and SEA AI Lab's Sailor2 sat side by side to talk about the infrastructure choices behind building large language models for Southeast Asia's languages and cultures. Connecting these local LLM efforts under one roof was, in many ways, the real headline of the day.
The energy didn't stop when the sun went down — the evening closed with "Dark Mode Devs," an open networking session that let the community keep the conversations going well past the formal agenda.
📖 Read the full recap and photos from the day on LinkedIn.
vLLM Meet Up Singapore: Keeping the Momentum Going
Singapore · 27 August 2025
After the scale of Developer Day, the community wanted something more intimate — a chance to reconvene on a regular weekday evening and go deeper on a smaller set of topics. That's exactly what the vLLM Meet Up Singapore delivered.
Held on a Wednesday evening, the meetup covered vLLM v1 development insights straight from a core contributor, an AMD deep dive on optimizing inference on data center GPUs, WEKA's Augmented Memory Grid for high-performance KV caching, and a practical session from the MERaLiON team on deploying AudioLLM at scale using Ray.
The event ended up oversubscribed — proof that four months after the inaugural Developer Day, appetite for vLLM knowledge-sharing in Singapore hadn't slowed down at all. If anything, the smaller, focused format showed that the community didn't just want big flagship events; they wanted a regular cadence to keep learning and building together.
vLLM Bangkok Meet Up: One Stage for Thailand's Local LLMs
Bangkok · 21 November 2025
From Singapore, the community's next stop was Thailand. The vLLM Bangkok Meet Up, held at GLOWFISH Sathorn on a Friday evening, brought together builders, researchers, and enthusiasts for an evening centered on one clear idea: give Thailand's own LLM ecosystem a shared stage.
The event featured a vLLM v1 deep dive from a core contributor, followed by lightning talks from the teams behind Typhoon and Pathumma LLM — two of Thailand's most prominent open-source language model projects — sharing how they're serving and training models tailored for the Thai language. A fireside chat let the local speakers dig deeper into where Thai AI development is headed next, while AMD and Red Hat rounded out the evening with sessions on ROCm software and the new vLLM Semantic Router.
Gathering Thailand's local LLM builders on a single stage — rather than scattered across separate meetups — was exactly the kind of connective moment the vLLM Asia community had set out to create back in Singapore.
vLLM Malaysia Day: Open Source Meets Enterprise
Kuala Lumpur · 2025
The year's community journey closed with its most ambitious gathering yet: vLLM Malaysia Day, hosted at Kuala Lumpur's iconic ILHAM Tower. If the earlier events were about bringing developers together, Malaysia Day was about bringing developers, enterprises, and policymakers into the same room.
The day opened with an official ceremony featuring MDEC, followed by a panel on cultivating deep tech and AI in Malaysia with leaders from AWS, YTL AI Labs, and 500 Global — squarely framing the event around Malaysia's push toward sovereign, production-grade AI as part of the ASEAN 2025 agenda.
From there, the day split into two parallel tracks. The Product Track covered technical ground — accelerating vLLM with LMCache, the vLLM Semantic Router, WEKA's approach to breaking the inference memory wall, and high-performance serving on AMD ROCm — alongside a hands-on workshop on building real-time agentic AI. The Enterprise Track ran alongside it, with sessions from Aras Integrasi, AMD, and an "AI MBA Fiesta" focused on what's actually working in AI for business today.
With over 500 attendees, vLLM Malaysia Day became the biggest single gathering of the year — a fitting way to close out a year that started with 300 people in a room in Singapore and ended with open-source communities, enterprises, and government all speaking the same language about AI inference.
Looking Back, Looking Ahead
Four events, three countries, and thousands of conversations later, one thing is clear: Asia's AI inference community isn't just growing — it's connecting. What began as a single evening of technical talks in Singapore became a recurring meetup, then a regional gathering of local LLM builders in Bangkok, and finally a full-scale summit bringing open-source and enterprise AI together in Malaysia.
If 2025 was the year this community found its footing, we can't wait to see where it goes next.
Want to be part of the next vLLM Asia event? Follow our LinkedIn page to be the first to know when registrations open — these events tend to fill up fast.