- Achieved a 25% reduction in storage usage and removed the data mart and warehouse complexity, without compromising high-frequency ECG data load performance (at 256Hz), by adopting Data Lakehouse architecture with strategic data partition strategy and Minio as the storage. This initiative significantly enhanced overall system efficiency, reduced maintenance overhead, and contributed to a more cost-effective and scalable data infrastructure.- Increased the training speed of the ML model by 40% and accuracy by 5% through the implementation of a robust cluster-based data sampling algorithm during pre-processing on highly imbalanced data. The notable impact of this enhancement is improved stability during the training phase, making a significant contribution to the overall model performance.- Accelerated AI development experience by 5x by designing and building a data platform automating migration & processing pipelines across diverse sources in a distributed manner, complete with a user-friendly Interface.- Improved the AI model confidence in detecting abnormal ECG patterns for patients by implementing a direct human feedback system integrated into the ML infrastructure, with comprehensive monitoring features called BeatLab.- Architected diverse purpose dashboards, including model monitoring and business-related metrics, resulting in improved data visualization, informed decision-making, and enhanced business strategy.

- Main developer of a modern predictive maintenance system (LUDA) which focuses on early detection of anomalies and abnormal activity detection in equipment operation in industries.- Collaborated with Korea Aerospace Industries (KAI) on Research and Development projects involving the implementation of machine learning system for real-time detection of anomalies and abnormalities with result of 88% prediction accuracy and 175ms latency.- Developed a real-time anomalies detection system using Spark Streaming + Kafka for high-frequency rotating machinery.- Implemented a high-performance Lakehouse architecture for an efficient data platform.Designed user dashboards for monitoring and Remaining Useful Life (RUL).