Meet the Startup Aiming to Reach the Top with Industrial AI
In August of last year, Gauss Labs Inc., the first artificial intelligence (AI) company of the SK Group, established its headquarters in Silicon Valley, USA, and cast its vote in the industrial AI market. The first challenge field of Gauss Labs is 'semiconductor', which is one of the biggest contributors to Korea's economic growth and is one of the most complex manufacturing processes. Among them, an AI solution that can solve the difficulties of SK hynix's semiconductor manufacturing process and enhance efficiency was selected as the first task.
It has been 10 months since Gauss Labs first started. Gauss Labs, which has been focusing on putting the core projects on the right track with a system in place, is preparing for another leap forward. It has been solidly built up to be able to promote quantitative growth.
Accordingly, by the end of this year, we plan to recruit new talents at twice the current size. Newsroom visited Gauss Labs to take a closer look at current projects and future possibilities, and to hear about the talents they want from management and the blueprints Gauss Labs is drawing on.
Gauss Labs aims to win the global industrial AI market with the start of K-Semiconductor
Unlike service AI, industrial AI does not have an overwhelming leader, so it is known as a 'blue ocean' with various opportunities. As such, the fields are diverse, and if you achieve results in any field, you can quickly acquire the position of a market leader. Among them, the field Gauss Labs selected was 'semiconductor', a decision made by combining the economic ripple effect created when AI is applied to the semiconductor industry and the needs of the SK group, which needed to strengthen the AI capabilities of its main affiliate, SK Hynix.
CEO Kim Young-han said, “The semiconductor industry is of the greatest importance worldwide and has a great impact on the Korean economy. In fact, in the industrial AI market, the semiconductor sector is expected to secure a value of 100 trillion won in the mid- to long-term.”
Above all, AI technology developed in the semiconductor process has the advantage that it is relatively easy to apply or spread to other industries . This is another reason why Gauss Labs, which was launched with the goal of becoming the No. 1 in the industrial AI market, chose the semiconductor industry as its starting point.
CEO Kim explained, “Semiconductor is called 'the flower of precision manufacturing, 'so it is an industry that combines various fields from chemical processes to mechanical processes and optical processes. He continued, “If the problem can be solved here, it will be possible to expand the scope to other fields. That is the goal of Gauss Labs,” he said.
Since its launch, Gauss Labs has been developing an AI solution that can maximize production efficiency by utilizing the vast amount of data generated at SK Hynix's manufacturing sites. Through this, SK hynix is promoting the intelligentization and optimization of the overall semiconductor production process, including process management, yield prediction, equipment maintenance, material measurement, defect inspection, and defect prevention.
Process monitoring solutions that combine AI to increase yields and lower costs
Gauss Labs is concentrating on developing a process monitoring solution using AI first to solve the challenges of the semiconductor process. Monitoring technology is an essential element of not only semiconductors, but all manufacturing processes in which raw materials are processed and converted into other forms.
In the manufacturing process, since reliability to ensure the same quality and robustness to minimize the impact of unexpected changes are important, it is necessary to check whether the product is being made properly. In particular, semiconductors are a field that requires precise manufacturing technology, and the tolerance for error is very low, and it is impossible to use it even a little bit outside of it.
“It is made up of hundreds of complicated processes, so there are a lot of defects . Therefore, engineers continuously monitor the status of the process or equipment in the fab (FAB) or in front of their PCs in real time. Gauss Labs is currently developing three solutions that utilize AI technology to enable engineers to perform monitoring tasks easily and accurately. It can be said that the purpose is not to replace engineers, but to create an AI Assistant that helps smart decision-making. ” - Kim Moo-Seong
In this regard, Gauss Labs is currently developing three solutions: △ Automatic Image Metrology for Semiconductor ( hereinafter referred to as AIMS), △ Virtual Metrology ( hereinafter referred to as VM), and △ Statistical Process Control ( hereinafter referred to as SPC).
AIMS is a solution that innovates metrology based on image data. In the semiconductor process, wafers are photographed with an electron microscope to find defects in wafers, and the original size and degree of damage are checked. The image data generated by SK hynix is an average of several million images per day.
AIMS improves image quality using computer vision (a technology that realizes the human visual system through a specific algorithm) and automatically performs measurements, enabling faster and more accurate measurement than before. Also, since measuring equipment is very expensive, cost reduction can be expected when AIMS is applied.
VM (Virtual Metrology) is a virtual metrology solution that can obtain measurement effects without actually measuring. It uses the sensor data of the equipment generated when wafers are processed in the equipment to predict and provide the required measurement values to engineers.
This is particularly useful for monitoring the condition of equipment and processes. If you apply the VM solution, you can reduce the investment in measuring equipment and maximize the efficiency of your current measuring equipment. The goal of Gauss Labs is to raise future forecasting power to a level that can replace actual measurement.
SPC (Statistical Process Control) is a monitoring solution that quickly informs engineers of the cause of a problem by analyzing the possibility of problems occurring in each process or equipment through machine learning when an abnormality occurs in the process. When this Root Cause Analysis technique is applied, simple repetitive tasks can be minimized, and equipment downtime (downtime) can be greatly reduced, and engineers can also help solve key issues. I can concentrate.
Gauss Labs expects that if the solution currently being developed is applied to the overall manufacturing process, it will create an economic ripple effect worth hundreds of billions of won.
“Ultimately, the goal is to build an industrial AI platform by converging the developed solutions . It will be difficult to apply the developed application itself to various industries as it is, but the underlying technologies constituting the app will be able to be easily applied to other industries as well . By building an open platform with reusable base technologies, SK hynix members as well as related companies such as semiconductor equipment , cloud and software will freely participate and develop an AI platform to achieve industrial AI innovation .” - Yoon Seong-hee
How does Gauss Labs work, an 'AI engineer navigating the manufacturing site'?
Gauss Labs is largely divided into an R&D organization that researches and develops AI solutions and a Program Management Office (PMO) organization that commercializes these AI solutions. Gauss Labs recruited the best experts in each field to strengthen the organization's expertise.
We hired Yoon Seong-hee from Amazon as the head of the R&D organization that will lead the technology development. Yoon Seong-hee is an AI and optimization expert who majored in Machine Learning and Convex Optimization and has accumulated skills in various industries such as semiconductors and e-commerce.
The PMO organization is led by Kim Moo-seong, a former engineer at SK Hynix and Intel. He has experience working as a citizen data scientist as well as a semiconductor process and equipment expert, so he is evaluated as an ambidextrous talent who is proficient in both fields of semiconductor and AI.
In addition, professors from leading universities in Korea in each field, from machine learning to statistics, control, vision, and scheduling, are also working part-time on the project. We are actively cooperating with academia, including participation in domestic and foreign conferences, and are planning to conduct research and human resource exchanges with leading universities in and out of the country in the near future.
The R&D organization, made up of engineers working on AI solutions, is subdivided into four teams according to technology. CV/IP team in charge of computer vision and image processing, ML/DS team in charge of machine learning and data science, OREO (Operations Research & Engineering Optimization) team that studies planning and scheduling of manufacturing processes, and software platform development SWAP (Software Architecture & Platform) is divided into teams.
They analyze problems based on their understanding of domain knowledge (professional knowledge) required to solve industrial AI problems, and research, develop, and develop AI algorithms using AI knowledge, intuition, and experience.
In addition, the R&D organization is also responsible for drawing up the mid / long-term technology roadmap of Gauss Labs based on the latest trends in AI technology in the industry.
A PMO organization is an organization that plans a product needed for a customer and manages a project to develop it. It should play the role of a communication medium that clearly defines the problems on the manufacturing floor and converts them into mathematical problems to AI engineers. Therefore, as a project manager and product owner, the PMO must set a big direction in the strategic aspect of the customer in the long term.
Due to the nature of Industrial AI problems, a single project often requires a convergence of multiple technologies. Therefore, Gauss Labs does not carry out independent projects for each organization or team, but rather gathers by project to create a synergy between members by performing tasks in a matrix structure. Maximize it.
In addition, Gauss Labs is closely collaborating with various related organizations such as SK Hynix Manufacturing/Technology, Future Technology Research Institute, and DT. contemplating together.
To this end, we share data and domain knowledge necessary for development, check that the solution being developed can be well settled in the actual field, and have regular meetings twice a week to discuss major projects together. To help Gauss Labs' AI engineers build domain knowledge in the semiconductor field, quarterly training, such as a Fab tour program, is also being conducted.
“People who know AI don't know the domain, and those who are in the domain don't know how AI works. You need to know both disciplines to effectively incorporate your domain knowledge into creating AI algorithms. However, for an organization of only AI engineers, it was determined that this would be difficult. Also, since Gauss Labs is not an internal organization of SK Hynix, it was difficult to collaborate closely. So, from the beginning of Gauss Labs, SK hynix formed a counterpart organization, and Gauss Labs is also working with SK hynix, a customer, as a One-Team. ” - Kim Moo-Seong
“We are looking for a 'Gaussian' to join us on our journey with Gauss Labs”
Currently, there are a total of 22 members working at Gauss Labs, and by the end of this year, it is planned to expand to 50 at the US headquarters and the Korean office. Recruitment in the United States, which had been delayed due to COVID-19, will start in earnest from June.
What competencies and qualifications do I need to have to work at Gauss Labs? Also, among many AI -related companies, why should it be Gauss Labs? Newsroom heard more about Gauss Labs' talent and vision from CEO Kim Young-han.
Q. What would be the appeal of AI work in the semiconductor field?
There is no other industry that accumulates as much data as semiconductors. Being able to directly deal with real data generated at the manufacturing site will be a valuable opportunity for AI engineers. In addition, the experience of directly dealing with real-world problems and data with members of the field is an experience that is difficult for startups to enjoy.
The semiconductor manufacturing field presents challenges that are difficult but worth the challenge. It is also attractive that it can create enormous economic impact by solving problems.
Q. What are the necessary competencies and qualifications for Gauss Labs engineers?
As Gauss Labs is walking on a path that no one has gone before, it must solve problems that existing AI technology cannot solve. We have to define what kind of problem we are going to solve by ourselves, and it is necessary to consider the insight that approaches the core of the question and the creativity and technology that can apply the algorithm in a new form. That is why we need engineers who can challenge various challenges and come up with creative solutions.
Gauss Labs is based on the 'Gaussian Principles' that define a total of 7 core values, including △ Focus on What and Why △ Highest Standards △ Learning with Growth Mindset △ Readiness for Deep Dive △ Ownership and Accountability △ Team Spirit △ Open Discussion and Communication. We prefer talents that match this. It is also used as an important evaluation criterion when conducting actual recruitment.
Q. What are the advantages of working at Gauss Labs as an AI engineer?
The main keywords that are currently the hottest topic in the world are by far semiconductors and AI. As such, semiconductors and AI are the most important technologies that determine national competitiveness and are the driving force behind the development of the world in the era of the 4th industrial revolution. Therefore, if Gauss Labs solves difficulties in the semiconductor field through AI and builds capabilities, it will be possible to grow into a key talent needed by the world
In addition, I think that it is a great advantage to be able to lay the groundwork to expand into various industries without being limited to the semiconductor area.
Q. Please leave a message for AI talents who will be with you in the future.
The industrial AI field is a blue ocean with no Google, Microsoft or Amazon yet. It's not an easy road, but it's also a place where you have a chance to be number one at any time.
Do not miss the opportunity to develop and grow the many AI knowledge you have learned in this vast ocean. At Gauss Labs, talented people with outstanding capabilities are currently taking on the challenge to become the best in industrial AI, and are waiting for a new Gaussian to join the journey in the future.
With Gauss Labs, you can learn a lot as an AI expert. We ask for a lot of support from AI talents.