Jeffrey Preston Bezos Executive Chairman of Amazon | Amazon
Amazon Web Services (AWS) has announced that Mesolitica, a Malaysian startup focused on training large language models (LLMs), has developed a Malaysian language generative AI model on AWS's cloud platform. The MaLLaM LLM is designed to understand local nuances such as slang and dialects, including Bahasa Malayu and 16 other regional languages. This model aims to enhance AI assistants in various industries by providing culturally relevant support for customer service, content generation, and data analysis.
Mesolitica achieved significant compute cost savings of 87% using custom machine learning chips like AWS Trainium and AWS Inferentia. This efficiency also resulted in a 5.5-fold increase in throughput during the training of MaLLaM. With AWS's resources, Mesolitica can deploy proofs of concept within 24 hours while benefiting from reduced latency provided by the AWS Asia Pacific (Malaysia) Region.
Dr. Kev Lim, CEO & Founder of Qmed Asia, stated: "Our biggest challenge was understanding the many local languages Malaysian patients use. By leaning into Mesolitica, we can now better understand local speech patterns through MaLLam."
The Malaysian government is exploring integrating MaLLaM into its operations to improve communication with citizens across different states and dialects. Khalil Nooh, co-founder and CEO of Mesolitica, emphasized the importance of this initiative: “With AWS, we can deploy proofs-of-concept much faster... making our MaLLaM generative AI assistant strategically important to the country’s digital transformation ambitions.”
Mesolitica has enhanced its machine learning operations by migrating workloads to Amazon Elastic Cloud Compute (EC2) and deploying inference workloads using EC2 G5 instances for GPU acceleration. The company also uses Amazon SageMaker for managing large datasets necessary for training LLMs.
As part of the AWS APJ Generative AI Spotlight program and receiving credits from the AWS Activate Program, Mesolitica is positioned to leverage AWS resources effectively for growth and success.