Mesolitica, a Malaysian startup focused on developing large language models (LLMs), has created a generative AI LLM designed for the Malaysian language, leveraging one of the world’s leading cloud platforms – AWS.
Amazon Web Services (AWS), an Amazon.com subsidiary focused on cloud services, announced that Mesolitica created an LLM for generative AI assistants that leverages the AI startup’s expertise in local language processing and AWS’s advanced cloud capabilities to understand Malaysian-specific nuances.
The model aims to improve customer service for small and medium enterprises (SMEs) across industries by automating tasks with natural language chatbots. To enable this, the culturally relevant LLM was trained to distinguish and respond to local language nuances like slang, colloquialisms that merge different dialects, Bahasa Malayu, and 16 other regional languages.
The solution tackles some of the main problems faced by businesses using AI pre-trained in English and western-centric data – cultural ignorance, inaccuracies and misinterpretations.
Dr. Kev Lim, CEO and Founder of health-tech startup Qmed Asia, explained that their biggest challenge was understanding the diverse local languages spoken by Malaysian patients. He noted that by leveraging Mesolitica’s MaLLam model, they were able to better comprehend local speech patterns. This improvement enhanced the accuracy of their medical note-taking system, strengthened communication with patients, and ultimately enabled them to provide higher-quality healthcare.
This innovation demonstrates the potential of generative AI to bridge digital gaps and streamline operations for various businesses in Malaysia and beyond.
Not only the private sector but also the Malaysian government is considering incorporating MaLLaM into its operations, supporting its goals of AI sovereignty and local data management. AI assistants powered by MaLLaM can deliver swift and precise responses to citizens in multiple languages, including regional dialects from states such as Johor, Kedah, Sarawak, Selangor, and Terengganu. This integration may enhance citizen engagement and improve data processing efficiency in Malaysia’s culturally diverse environment.
Additionally, the educational sector could leverage MaLLaM for language learning and research, aiding in the study and preservation of local languages and dialects.
Mesolitica has chosen AWS as its partner in the new project to achieve compute cost savings and an increase in throughput (transactions per second). Khalil Nooh, co-founder and CEO of Mesolitica, explained that using AWS has significantly accelerated their ability to deploy proofs-of-concept by providing cost-effective AI computing resources and advanced machine learning capabilities. He emphasized that this approach allows clients to focus on manageable operational expenses rather than large upfront investments in AI experimentation.
Mesolitica optimized its machine learning operations by using AWS services like Amazon EC2, Amazon EKS, and Amazon SageMaker. These tools enable efficient training, deployment, and management of AI models while reducing costs through GPU-accelerated infrastructure. Besides, the startup is leveraging other benefits of AWS support available within the AWS APJ Generative AI Spotlight accelerator program, where it participates.
