Co-organised by Amazon Web Services (AWS) with AI Singapore (AISG), the first-of-its-kind Regional LLM League drew 1,300 student participants from higher-education institutions across six countries: Indonesia, Malaysia, Philippines, Singapore, Thailand and Vietnam. The competition kicked off in January, culminating in the Grand Finale held in May as part of the AI Student Developer Conference (AISDC). Organised by AI Singapore (AISG), AISDC brings together industry leaders and emerging AI talent to drive innovation and shape the future of technology. The conference showcases real-world AI applications, career opportunities, and thought leadership through panel discussions, workshops, and exhibitions. It aims to inspire and equip the next generation of AI talent in Singapore and the region.
“LLM League began as a bold idea from the fearless innovators at AWS – to gamify LLM fine-tuning and make it engaging for all,” Mr Koo Sengmeng, Director, AI Talent Development, AI Singapore, said. “AI Singapore is proud to partner with AWS to bring this essential 21st-century skill to life in a fun and accessible way, not just in Singapore but across the ASEAN region.”
Elsie Tan, Country Manager, Worldwide Public Sector for AWS Singapore, said: “The LLM League embodies our vision of digital inclusion for learners from all backgrounds, including students. I am excited we were able to bring together so many talented individuals to master cutting-edge AI technologies. To date, AWS has trained 31 million learners across 200 countries and territories build their cloud skills through its free training initiatives, creating pathways for underserved communities to shape tomorrow’s technology.”
A showcase of student creativity and innovation
To compete, participants were given Amazon SageMaker JumpStart credits to fine-tune Llama 3B models, with the end goal of challenging a much larger 70B reference model and outperforming it in a quiz-based evaluation. Through an intensive Generative AI (Gen AI) workshop conducted by facilitators from Gen-C, AWS’s Gen AI Learning Community, the students were introduced to core concepts in LLM development, including dataset curation, prompt engineering and evaluation techniques, as well as agentic and responsible AI. In just three weeks, they created over 5,000 models – demonstrating both the effectiveness of the program, and the value of skill and strategic fine-tuning over size when it comes to AI models.
Following a series of local competition rounds and internal evaluations, the top-performing students from the six countries were then invited to participate in the Regional LLM League finals in-person at Equarius Hotel in Singapore on 29 May. Each student was assessed in two key areas: their model’s accuracy in quiz-based question answering, and their ability to carry out prompt tuning in real-time to optimise the model’s responses. The judging panel included experts from AISG and AWS, audience voters – and, of course, AI itself, via a benchmark powered by Anthropic Claude 3.5 Sonnet to ensure a fair and comprehensive evaluation process.
In the end, computer science student Blix D. Foryasen from the Philippines emerged in first place. Through model distillation – using the outputs from a larger ‘teacher’ model to train a smaller model – and experimenting with various teacher models to generate high-quality instruction data, his model defeated a larger one in the quiz.
“Evaluating model responses in dataset curation and inference generation was my biggest, most enjoyable takeaway,” Foryasen said. “This challenge revealed AI’s capabilities and limits, proving that human evaluation remains crucial for bias mitigation and reducing erroneous outputs, even with the increasing popularity of AI agents.”
In second place was Vietnam’s Kim Seokyung, a software engineering student who was experimenting with LLM fine-tuning for the first time – showing just how accessible the learning journey can be. “I went from zero to a fine-tuned model ready for real use in just weeks,” she said. “It was intense but rewarding, and I’m now confident applying this skill to real-world use cases in growing demand across industries.”
Building individuals’ skills for the public good
One key focus of the Regional LLM League was hands-on learning in a real-world landscape – the development scenarios that participants had to tackle mimicked those of a professional AI environment. The students were also provided with industry-grade Amazon SageMaker JumpStart tools for model training, inference testing and deployment simulation. This arrangement helped them build up both their technical proficiency as well as their familiarity with the field.
The success of this first regional tournament signals the potential for a scalable, community-led model of technical upskilling. As a program under the AI Spring Communities pillar under AWS AI Spring Singapore, it serves as a blueprint for how AI education can be democratised at scale, especially in regions with diverse education systems and access levels, and across underserved and underrepresented communities. By embedding Gen AI education in challenge-based formats and tying it to community impact goals, initiatives like this bridge the gap between curiosity and a career in AI. Students leave with a portfolio of real projects, experience working with small and large models, and the confidence to pursue internships, research roles, and start their own AI ventures.
As Gen AI becomes a central pillar of digital transformation, the need for a diverse, inclusive talent pipeline is more important than ever. The Regional LLM League stands as proof that, with the right tools, mentorship and opportunities, anyone can contribute positively to the future of AI development.
Learn more about how AWS is investing in education, workforce development and innovation for the public good here.
