At Samsung Research, we use numerous cutting-edge techniques to extract insight from various data sources, data types, and use cases. Here are just a few of them.

Knowledge graphs (KG) are important in information representation and decision making. As such, we have been developing component technologies and platforms for systematically building, maintaining, and effectively utilizing them. Our research topics include ML-based models for automated knowledge graph construction from different data sources, neural / logic-based reasoning engines, KG embedding (KGE) models and associated KGE-based downstream ML models, models for enabling KG-based semantic search with or without recently emerging large language models, and various optimization techniques to make all of these scalable and computationally efficient at both engine and platform level. Our technologies are helping to make our AI products (Bixby, SmartThings, future robot devices, etc.) more intelligent and proactive to better cope with each user’s context. Moreover, our ultimate vision is to build a unified Samsung graph to enhance our services, businesses, and management by seamlessly utilizing the data collected in the vast Samsung ecosystem.

We actively develop models and algorithms for optimizing ads and other market-based products. Research items include, but are not limited to, graph NNs (GNNs) for supporting scalable integrated data analysis of heterogeneous data sources, multi–task learning for efficiently training the task relationship in a multi–objective recommendation, domain adaptive/transferrable/interoperable modeling for rapidly increasing cold-start ads campaigns, and reinforcement learning for optimizing multiple business key performance indicators (KPIs) harmoniously.

In addition, we are conducting digital health-care data research, such as data imputation and anomaly detection technology to process time-series health sensor data, and developing recommendation/coaching systems to provide personalized well-being health-care services in exercise, diet, sleep, and mindfulness.

The Data Cloud team focuses on hardware (HW) infrastructure and software (SW) solutions for large-scale, complex data management, including high-quality data acquisition, massive data storage / repository management, and business analytics tools. We custom design the system for our researchers’ needs and secure the cost-effective infrastructure needed to acquire and manage the large-scale datasets utilized by our scientists. Our data cloud technologies provide a competitive advantage to our organization by providing our researchers with a fast and modern environment to extract value from Samsung’s vast data resources.

Our AI infrastructure research and deployment focuses on developing a world-class AI development environment with integrated physical HW and advanced SW systems that provide complete end-to-end ML operations ranging from data processing to AI model deployment. We are developing and deploying autonomous and scalable deep learning technologies based on the latest cloud technology. Ultimately, we seek to provide our AI platform and infrastructure to all Samsung AI developers and build an ecosystem enabling them to collaborate and share to leverage each other’s experiences to grow and be more productive.