vLLM Bangkok Day: the creators are coming to Bangkok
Something that has not happened before in Thailand is happening on Sunday 6 September 2026: the core maintainers of vLLM are coming to Bangkok.
vLLM Bangkok Day runs from 14:00 to 19:00 at JAS Green Village, Ramkhamhaeng. It is free, and it is aimed squarely at people who run — or want to run — large language models in production.
Why this matters for Thailand
vLLM is the open-source inference engine behind a large share of the world’s LLM serving. If you have deployed an open-weights model behind an OpenAI-compatible endpoint in the last two years, there is a good chance vLLM was doing the work.
That matters here for a specific reason. Thai teams building on open models are making the same decisions — batching strategy, KV cache sizing, quantisation, GPU spend — mostly in isolation, and mostly by reading GitHub issues at midnight. This event exists to shorten that loop:
- Ask the maintainers the question directly, instead of filing an issue and waiting.
- Hear what Thai teams already running vLLM in production got wrong, before you get it wrong too.
- Find out who else in Bangkok is solving the problem you are solving this quarter.
What is on
Talks by the vLLM creators. Maintainers from the USA and Singapore on the internals — paged attention, continuous batching, and where the project is going next.
Real production use cases. Thai companies, developers and researchers on what actually broke, what it cost, and what they would do differently. Not launch announcements — post-mortems.
Enterprise and technical sessions. Deployment patterns, GPU economics, and the specific problem of serving Thai-language models at production latency.
Networking, Thailand ↔ global. Structured time to meet the international vLLM community and the people building AI infrastructure here.
Who should come
You will get the most out of this if you are:
- serving an open-weights model today and want it faster or cheaper
- evaluating whether to self-host instead of paying per token
- researching inference, quantisation or serving systems
- responsible for the GPU bill
You do not need to be a vLLM contributor. You do not need to be deep in CUDA. Bring the problem you are actually stuck on.
Practical details
| Date | Sunday 6 September 2026 |
| Time | 14:00 – 19:00 (ICT, UTC+7) |
| Venue | JAS Green Village, Ramkhamhaeng, Bangkok |
| Cost | Free — registration required |
| Language | English and Thai |
Over 200 Thai and international attendees are expected. The countdown on the event page tells you exactly how long you have.
Get involved
If your team runs vLLM in Thailand, add yourself to the community directory. We are building a picture of who is actually doing this work here — and it is more people than any of us realised.
If your company wants to support the event, the sponsorship packages are open.
See you on 6 September.