I am currently serving as a team lead for the advanced research and development unit at AWL, Inc. My research at AWL focuses on enhancing the accuracy and performance of computer vision models, particularly detection and classification models. I am also involved in improving and automating the AI pipeline that supports model lifecycle management (through continuous fine-tuning). In addition to development, I actively supervise my team in research and implementation tasks.

During my personal time, I engage in projects related to modular neural networks. This includes training an ensemble of expert neural networks and exploring innovative ideas for smarter design, such as data partitioning and clever subset class partitioning in ensemble machine learning models. I am working on implementing a large ensemble of learning models where only a portion of them will be activated during inference. This approach resembles the Mixture-of-Experts (MoE) approach.

Before my current role, I pursued a Ph.D. degree at the University of Aizu in Japan, under the supervision of Prof. Qiangfu Zhao (https://u-aizu.ac.jp/~qf-zhao/) with a focus on knowledge distillation for model compression and the architecture modularization of deep neural networks, specifically employing a mixture of independent experts for resource-efficient training and inference.