Startup Interview

  • [Startup Interview] Kyungwon Oh, CEO of ‘CARILLON’, a company servicing ‘Legal Engine’ for legal professionals based on AI Legal Research
  • 창업원
  • 2022-10-30 20:22:17
  • 226

Carillon, winner of 2020 Fall E*5 KAIST, joins us for Startup Interview! Carillon provides a legal search service called ‘Legal Engine’ that helps legal professionals by digitalizing legal data based on AI.

Carillon is a legal tech company that started with a lawyer and an engineer with its specialized machine learning technology. Let’s meet with CEO Kyungwon Oh.

Can you briefly introduce Carillon?

Carillon is a company servicing ‘Legal Engine’ (, a legal search engine that helps finds solutions. Using AI technologies like OCR, data anonymization, legal machine translation (BLEU Score 59), Carillon digitalizes legal documents and provides 3.3 million legal data. You can find data from various institutions such as the Fair Trade Commission, the Financial Services Commission, and the Tax Tribunal at once, and check the English translation immediately.

The company name is unique. Why did you name the company ‘Carillon’?

Carillon has various historical meanings, and is also important in law. The symbol of Seoul National University’s law school is “the Bell of Justice”. Pyeonjong, also known as Carillon, is also representative of KAIST. Since we are KAIST-born entrepreneurs in the legal market, we named our company ‘Carillon’ to innovate the world by connecting law and engineering.

I didn’t know the name of the bell at KAIST was Carillon!

How did Carillon decide on this service as its startup item?

The legal research service started from my litigation experience. I received a lot of help from my lawyer friends and in the process, I found out that lawyers work in a difficult environment.

For example, the court library is the only place in Korea where all rulings can be accessed. Furthermore, the library is by reservation only, and you cannot download or print resources. The legal field does not utilize a lot of technology, instead using hard copies that cannot be searched online. I was confident that I could innovate the field with AI and started my own business.

The lack of a platform for legal professionals to share rulings is surprising. Lawyers must have been spending a lot of time researching rulings, since resources cannot be downloaded from the library. I believe Legal Research using AI technology can solve the difficulties that legal professionals face.

What exactly does a legal service for legal professional mean? Is it also provided for the general public?

Legal Research is the process of finding an answer to a legal question, and becomes the most fundamental work of all legal affairs. The function of legal research is to find out how previous courts and institutions have decided with similar cases. Thus, it is the process of researching the outcome when there is a legal problem.

Non-legal professionals can also use legal research. However, it is difficult for non-legal professionals to link legal questions to correct principles. In addition, even if they do find similar rulings, it is difficult to clearly interpret the results. So, we provide a “post” that summarizes rulings related to major issues for the public. For example, a lawyer analyzed and provided previous cases that received a verdict of guilty and not guilty related to the “Minsik Law” (

In the future, we will apply natural language search technology to increase the “right for trial” for people rather than the current keyword search engine.

I looked around the Legal Engine website during the interview, and I think it resolved my curiosity on rulings and is easy to understand. If the service is provided to the public in addition to legal professionals, it will help approach cases with similar fact patterns.

What is unique about Carillon from other existing legal tech service companies?

The biggest difference is that we are a well-balanced team of lawyers and engineers. Carillon consists of lawyers who fought for more than 300 billion won a year, and doctors of engineering. With this talented workforce, we could clearly define and implement the needs of legal professionals.

In addition, unlike other legal tech services that are based on labor, Carillon actively uses machine learning technology. For example, the legal translator that we developed and are currently providing the alpha version is ahead of Google translator in the BLEU Score (indicator of translation quality). Furthermore, Carillon develops its own model such as OCR, spell check, anonymization, to provide the best legal data.

Since Carillon is a team of lawyers and engineers, the company will make a more applicable and useful service for legal professionals. AI and developing its own model will become meaningful to develop legal services.

What motivated you to participate in E*5 KAIST, even though you had already started the business? What did you learn from E*5 KAIST?

Many alumni who participated in E*5 KAIST in the past recommended the program. I became interested in the program, and decided to participate after seeing the program details. In the beginning of the program, we had already established a corporation and launched beta services. The sharp feedback from mentors and judges became extremely helpful to our business. Since mentors were entrepreneurs from engineering, their advice became more practical and helpful.

Even though Carillon had already launched its service, the feedback from mentors and experts became helpful. Mentors and judges are from the startup ecosystem, so they must have analyzed the business more objectively than anyone else. I hope it was a helpful time to decide on the future business direction.

What were some difficulties starting a business, and how did you overcome them?

The most difficult part of starting a business is that new problems appear every day, and I must make optimal decisions immediately. From deciding on the color of the button in the service to hiring an interviewee, I must make fast decisions. I make most of the decisions, but regarding decisions that impact the whole company, I consult with team members. In the process, I realize what I missed and organize my thoughts.

Since you participated in E*5 KAIST after you started the business, you must have experienced the problems in the field directly. Making decisions as a team shows that Carillon is a company that communicates well.

What is Carillon developing next?

Our ultimate service that we are aiming for is a “Mind reading legal search engine service”. To accomplish this, we need to develop natural language search. For example, when we get a question like “I suffered from noise between floors for 3 years. How much can I get for compensation for damages?”, we need to find appropriate data related to this problem.

A mind reading legal search service seems open to the public! The process of receiving legal advice and finding related data will become much easier.

What is Carillon’s ultimate goal?

We have two goals. First, we want to continue to help legal professionals work efficiently using AI technology. Ultimately, we want to create a legal market where legal professionals can represent more customers, and anyone can get better legal support.

Secondly, we want to become the best tech company with text-related machine learning technology. We are currently dealing with legal documents using many AI models, but there are still many things to develop and solve. Carillon’s goal is to solve these problems well.

If existing legal tech companies approached the service in the client’s perspective, Carillon provides services for legal professionals. The legal service for clients is also important, but it is important to create a platform where professionals can search and share legal data and rulings. With Carillon’s Legal Engine, we expect legal professionals to work efficiently and solve legal problems more quickly. We hope Carillon becomes a company that can provide legal help to anyone in the market with its AI technology.

This was CEO Kyungwon Oh of Carillon.