NVIDIA AI Day Singapore opened Sept. 22-23 at the Raffles City Convention Centre with a concrete agenda: helping governments and enterprises across Southeast Asia move AI from pilots to production. The event features hands-on training, expert-led sessions and advanced tools, alongside announcements of region-specific deployments spanning public safety, finance, legal tech and smart cities.
Four national AI programs and one new hardware platform
Singapore's HTX (Home Team Science and Technology Agency) is researching the NVIDIA Nemotron 3 Super and Nemotron 3 Nano Omni models for public safety. Super targets complex reasoning and agentic workflows, while Omni's unified vision, audio and language capabilities are meant to support multimodal applications built on operational data. NCS is pushing agentic AI into enterprises and the public sector using Nemotron models and the NVIDIA Blueprint for video search and summarization (VSS), alongside humanoid robotics work. ST Engineering is applying NVIDIA NeMo tools and NVIDIA cuOpt software to build its AI Studio platform and deploy agentic AI in businesses such as Marine MRO.
The mechanics behind these deployments rest on open models that organizations can customize and control. In Thailand, iApp Technology, a member of the ThaiLLM Collaboration with the Big Data Institute, fine-tuned Nemotron 3 Nano into OpenThai 2.0 Legal using Thai-language legal data through the NVIDIA NeMo framework. The model was released as open source and now powers Thanoy, a legal-assistant chatbot serving approximately 43,000 users. Malaysia's YTL AI Labs and Vietnam's Viettel AI are fine-tuning Nemotron for enterprise, citizen and Vietnamese-language services, while Brunei's Antrique built an AI innovation platform for the food sector. For smart cities, Malaysia's ITMAX combines NVIDIA Cosmos world models with the VSS blueprint for traffic operations, and Thailand's AS-TECH applies the same stack to airport passenger flow.
The regional push follows a broader shift: governments now treat AI as national strategy and demand locally controlled systems reflecting local languages and cultures. NVIDIA Nemotron addresses this with persona datasets containing synthetic data grounded in regional populations and workforces. AI Singapore is extending its SEA-LION model family with Nemotron open models and NeMo tools. Viettel AI's fine-tuned Nemotron 3 Super took the top spot on the VMLU benchmark and the company's internal product benchmark, and will power a Legal AI agent harness for internal and external users. FPT Smart Cloud codeveloped Nemotron-Personas-Vietnam, an open dataset built on Vietnamese demographic data. Hummingbird Bioscience and LynxKite are building an explainable Toxicity Knowledge Graph on Nemotron 3.5 Lightning and NeMo Retriever to speed up drug discovery, while Bitdeer AI co-hosted an Open Models AI Codefest with NVIDIA, supplying GPU cloud infrastructure for developers.
What this means for companies adopting AI
The most visible commercial signal is hardware: Sea Limited, which runs Garena, Monee and Shopee, is the first enterprise in the ASEAN region to adopt the NVIDIA Vera Rubin platform. On Shopee, AI already shortens the work of creating product listings and improves product discovery; at Monee it supports fraud detection and credit risk assessment. For small and mid-sized sellers in Southeast Asia, this puts enterprise-grade capabilities within reach that were previously limited to large corporations. The pattern generalizes: open regional models plus partner-delivered infrastructure let mid-size firms adopt AI without building everything in-house.
What still needs verification is how much of the announced work moves past the pilot stage. Several projects are described as research or exploration — HTX's work, the ThaiLLM Collaboration's foundation-model study, Antrique's platform — and production timelines are not stated. The Nemotron-based models vary in maturity: Thanoy already serves about 43,000 users, while others are at fine-tuning or benchmark stage. Buyers should ask vendors which of these deployments are live, what data governance applies, and whether the open-source licensing of adapted models permits commercial use.
The clearest marker to watch is Sea Limited's Vera Rubin rollout across Shopee, Monee and Garena: if the first ASEAN deployment reaches production workloads at scale, it confirms that region-specialized open models and new NVIDIA hardware can carry real traffic rather than remain showcase projects.
