headerSearch form

Changing the World through Creative Research

AI Algorithms

We conduct fundamental research on various machine learning algorithms as well as computer vision and language understanding, and apply them to build industrial AI solutions in the fields of semiconductor manufacturing, materials discovery, and autonomous machine.

AI Algorithms

Understanding users and environments with human-level intelligence

AI Algorithms

AI for Semiconductor Manufacturing We are transforming semiconductor R&D and manufacturing processes using cutting-edge AI technologies such as
deep learning and generative AI. This AI-based transformation is creating fundamental changes across a range of areas,
from chip design to process development, yield optimization, metrology & inspection, fab monitoring and automation,
as well as customer and business environment analysis and decision making.

AI for Semiconductor Manufacturing

LLM Large Language Model (LLM) is a key inference engine that can be used in all semiconductor manufacturing and development fields by learning semiconductor knowledge and reasoning capabilities from various text data such as semiconductor reports, papers, patents, technical reports, etc., as well as general text data. It aims to assist semiconductor engineers and automate the entire manufacturing process.

MMFMA Multi-Modal Foundation Model (MMFM) is an advanced type of artificial intelligence that can process and integrate multiple type of data, such as documents, images, text, chart, table, and sensor data. These models are useful for semiconductor manufacturing and development, because they can help with tasks such as defect reasoning, quality control, design optimization and data analysis.

Customer & Enterprise PlanningAI innovation is needed to support critical decision-making in sales and customer response. AI can analyze customer quality data to prioritize response strategies and efficiently handle tasks related to customer issues. It also can analyze knowledge about the semiconductor industry to identify demand patterns and establish marketing strategies by predicting future purchase demand.

DesignSemiconductor design is a complex and time-consuming task that requires significant investment, and AI technology can maximize its efficiency. AI can quickly analyze large volumes of design data, optimize and solve complex problems that arise during the design process, and perform tasks such as reproducing defects through problem analysis on a scale beyond human limitations. This can greatly improve productivity in the fast-paced industry of design and enhance product quality.

ProcessWe maximize the efficiency of semiconductor R&D by applying AI technology. By identifying and optimizing potential problems throughout the process, we can reduce the lead time of process development. And by developing AI models that reflect the physical and chemical knowledge of the process, we can improve the performance of the final product even with small sample data.

Metrology & InspectionTo measure and inspect the increasingly complex and smaller semiconductor structures, AI models are utilizing various inputs such as Scanning Electron Microscopy, Optic Microscopy, Atomic Force Microscopy, X-Ray, Ellipsometry, and Time-series Data to predict 3D structures and detect defects. Ultimately, AI technology is utilized to predict structures and defects using all data from the semiconductor manufacturing process, control factors causing defects, and improve the process.

Manufacturing (Yield analysis, Anomaly detection, Optimization)We apply deep learning to semiconductor manufacturing for automated detection of potential faults, process optimization and yield prediction & analysis to maximize manufacturing performance. Our research topics in semiconductor manufacturing include Machine Learning, Deep Learning, Anomaly Detection, Explainable AI, Bayesian Optimization, Large Language Model, and Data Engineering/Science.

Data ManagementData management is one of the toughest hurdles in bringing advanced AI algorithms to the industry. Working closely with experts in the field, we are actively developing technologies to enable AI to handle every aspect of the data process, from collection and cleansing to processing and usage, by AI themselves.

FMOpsEntering the foundation model era, we are making notable changes in operating functions essential for efficiently managing data, aligning, deploying, optimizing, and monitoring large models within the framework of AI systems. In particular, we are working on building a data pipeline and SW platform (FMOps) that can efficiently develop and utilize semiconductor-specific models using various and large unstructured data.

AI for Materials Discovery Our materials discovery AI research aims to develop technologies that can assist in developing materials across various
industrial sectors by leveraging AI technologies. The technologies may include methods to sense new technology,
methods to explore and optimize the molecular structure and synthesis recipe of materials with desired properties, and more.
These technologies can play a major role in many areas, not only supporting researchers to efficiently complete various
tasks required for materials discovery, but also helping them make decisions to effectively explore the search space for
complex and wide-ranging optimization problems in the field of materials discovery.

We are advancing AI technologies through collaboration with high-priority materials discovery projects within the Samsung
group, aiming to meet domain-specific requirements and directly achieve our materials development goals.

AI for Materials Discovery

Display Material The advancement of display technology necessitates the research and development of cutting-edge materials like OLED and
QD (Quantum Dots). These materials play a vital role in maximizing display performance and enhancing user experience.
The application of Artificial Intelligence (AI) in this field is bringing about revolutionary changes, where AI serves as a critical
tool in analyzing complex data and predicting the physical and chemical properties of materials. AI contributes to exploring
new material combinations, optimizing performance, and increasing the efficiency of production processes.

In the development of OLED and QD materials, the role of AI is becoming increasingly significant. In addition,
AI leverages large datasets to learn the complex relationships between the characteristics and performance of materials,
supporting materials researchers in designing and optimizing target materials more quickly and accurately. This accelerates
the development cycle and opens up possibilities for introducing innovative display materials to the market more rapidly.
Such applications of AI mark a pivotal turning point in shaping the future of material discovery for the display domain.

Semiconductor Material The semiconductor industry is a field that requires continuous technological innovation, and the development of
materials/process technologies is one of the key elements of this innovation. With the recent combination of AI technology
with chemical and materials science, the use of AI technology is also playing a very important role in research and
development processes to explore materials and improve process/analysis technologies.

Based on advanced data analysis, the AI model for predicting experimental results and proposing experimental conditions
provides the potential to efficiently reduce the cost of research and development. In particular, the AI model can quickly
derive optimal material combinations and process methods by considering numerous chemical and physical properties,
which can accelerate the development speed of high-performance low-power semiconductors. In addition, it improves the
efficiency of the research process through risk management in experiments and contributes to allowing material
development researchers to focus on advanced research and development processes. As such, the integration of AI
technology is becoming an important driving force to innovatively change the future of the material discovery for
semiconductors.

We are focusing on developing AI technology to develop various materials and process optimization to enhance our semiconductor technology competitiveness and implement next-generation semiconductors.

Autonomous Machine AI

Semiconductor Material

We will secure Samsung‘s future technological competitiveness by realizing Autonomous Machine technology that
perceives, judges, and acts like humans through Embodied AI technology, which represents the next state of AI innovation
following Generative AI and Large Language Models.

Embodied AI We are pioneering the development of Embodied AI technology, a groundbreaking innovation that not only operates
autonomously with its own physical form, but also learns and adapts through dynamic interaction with its environment.
This technology harnesses the power of human-like cognitive abilities to make independent decisions. The core of
Embodied AI consists of situation awareness intelligence, which comprehends the causality and dynamics of the physical
world utilizing data from diverse modalities at a human-like level, and motion intelligence, capable of executing precise
machine control whilst adapting seamlessly to ever-changing environments.

Cognitive Intelligence / Decision-making Process We are developing embodied AI technology that perceives and makes decisions like humans. Unlike traditional robot
technologies that focus on designing robot behaviors by recognizing types of objects and spatial information in the
environment, our embodied AI technology predicts the actions that robots should take based on an understanding of
physical phenomena, object properties, and human-like reasoning. This enables to solve complex and diverse problems that
are difficult for humans to design and implement autonomous decision-making AI technology.

Motion Intelligence We are developing AI technologies related to the motion of future robots and mobility. Utilizing Embodied AI technology,
we create robot-specific motion AI models that perform complex and diverse tasks in real-world scenarios with greater
speed and efficiency.

These models foster intelligent, adaptive, and interactive capabilities, ensuring that our autonomous machines possess
cutting-edge motion intelligence. Through these advancements, we secure the core technologies essential for the future of
autonomous robotics.