The development of AI systems has been model-centric for decades. We consider modeling to be the heavy-lifting part requiring deep expertise to get it right. And the data preparation part is a tedious yet trivial job to complete as part of the pipeline. Although ensuring the data quality is understood to be a vital part of the pipeline, AI system developers often consider this part static. Once the data has been collected and prepared, we dedicate most of our attention to creating an accurate model. Data-centric AI challenges this concept by showing that effort spent on improving the data quality can also significantly enhance the performance of your system.

We design our platform with the data-centric AI principle in mind. With data cleaning and error analysis tools available on the VISAI AI Platform, AI developers can inspect the input data and results to improve the system performance.

The main technical barrier in AI system development is the application of data to a model. The process requires machine learning knowledge to find the best model settings for each application. The VISAI AI Platform tackles this challenge by streamlining this step. Our user interfaces hide the technical details of model parameter adjustments freeing AI developers from the model development part of the process while ensuring the best system performance.

At the core of our platform, we provide a comprehensive collection of base AI models that can be adapted to one's business application. Based on our experience as a research center developing the basic AI infrastructure of the nation, we have created high-performance models covering basic tasks in natural language processing. In particular, we have made publicly available models such as machine translation, named-entity recognition, automatic speech recognition, speech emotion recognition, and WangchanBERTa, which is the most advanced Thai Language Model up-to-date. In addition, we have also developed proprietary models covering basic tasks in data analytics and computer vision. Users of our platform will be able to apply these models to the ML automation functionality and develop AI solutions using our data-centric AI interfaces.