KAIST StartupTing X Minds Lab Review (May)

The second StartupTing of May was joined by Minds Lab. Minds Lab is a famous AI company engaged in various businesses from algorithms to platform and AI-related services. There are many AI experts as well as students from KAIST at Minds Lab. Through the stories of relatable KAIST seniors, StartupTing with Minds Lab was a time to hear about realistic stories about the company, team, and startup experience.

How do you become an AI researcher- Brain Team Researcher Hyoungkyu Song

Researcher Hyoungkyu Song of the Brain Team is working at Minds Lab as a skilled industry personnel. Researcher Song double majored in Bio and Brain Engineering and Computer Science and participated in various school activities at KAIST. The story of researcher Song will be described in a first person point of view.

Unlike other schools, KAIST has a lot of club and activities available, so I recommend students to work on various projects and try out intern. In my case, during my first internship at Nota, I wrote work shop papers, and was able to participate in various projects in the company. For deep learning tasks, one can only grow and gain skills through working on projects. To do deep learning, there are many requirements like datasets and servers, and Minds Lab provides these resources such as GPU. In the Brain team that I belong to, there are people from various departments and we also collaborate with many departments. I am in charge of audio and graphics, and develop and tune parts that are used in services like TTS and chatbots. At the Brain team at Minds Lab, we read research papers, implement them, and upload the code to paperswithcode. We also do paper review and participate in various conferences (CVPR, ECCV, ICLR). Most importantly, I recommend interning or working at a startup because you can service the engines you develop as APIs.

From a non-major to an AI Scientist – Brain Vision Team Researcher Seung min Yang

The second lecture was by Researcher Seung min Yang of the Brain Vision Team. Researcher Yang shared his feelings while working at Minds Lab for about a year. Researcher Seung min majored in Bio and Brain Engineering, minored in Biological Sciences, and is currently at Minds Lab as a skilled industry personnel as well. The story of researcher Yang will be described in a first person point of view.

In the beginning, I lacked background knowledge of computer science. I also participated in various activities, one of which is the Cuop program, where I worked at a healthcare startup. Based on my experience, I will explain how I became a researcher, the skills AI researchers need, and what I felt about Minds Lab as a researcher. While attending seminars in various fields, I became interested in artificial intelligence, but had a lot of worries about my career because graduation was ahead. In addition, I felt my lack of basic knowledge in artificial intelligence research. After joining Minds Lab and sharing knowledge with team members, I gained the knowledge required in the field and in work. The research I conducted are related to service development.

– Skin property Classification: Skin Type, Humidity, Symptom

– Comics OCR: Optical Character Recognition to transform Webtoons into Web novels

– Clothes Segmentation &Warping (Virtual TryOn): User takes a picture of a clothing, AI removes the background and tres the virtual outfit to an avatar

The skills required as an AI researcher are basic knowledge of computer science. One must understand the infrastructure to train AI (server, GPU, etc.), algorithms, data structures (especially tensors), programming languages used in research development, and an understanding of frequently used libraries (*Python – pytorch, pillow, opencv, torchvision, etc.). It is also important to do paper review, learn SOTA Algorithm and methods, understand an edit Open Source Repository, and implement unpublished source codes. Last but not least, communication is extremely important. Within the team, it is important to exchange and acquire background knowledge based on computer/NLP/Audio and data domain divisions as well as new insights. As a researcher, this is how I felt about Minds Lab. – Infrastructure: a lot of infrastructure and resources that researchers can use, and a lot of freedom in research. – Various fields of study: high interaction with other teams, learning beyond work, chance to work in other fields – Positive work environment: flexible commuting, encouraging participation in global conferences. Thus, I recommend working at startups like Minds Lab.

How do AI Scientists work? – Brain Audio Team Researcher Wonbin Jung

The last lecturer was researcher Wonbin Jung, who joined the company most recently. Researcher Wonbin Jung is currently at the Brain Audio Team, and double majored in Computer Science and Industrial and Systems Engineering at KAIST. He is a new employee, who has been working as a skilled industry personnel since March. At the Brain Audio team, Researcher Wonbin Jung is in charge of digitalizing analog signals. Brain Audio: TTS (Twitch) / AI Mouth: Conditional TTS / Brain team: Algorithm session / Code session / General Research – To work on these, it is necessary to understand the concepts and theories of deep learning and directly experience implementation methods (Github, Python, etc.). Interest, curiosity, skills in English, and Googling become helpful assets to a researcher. Researcher Wonbin Jung shared his 3-month experience at Minds Lab in detail.

Q&A and Discussion

The Q&A session was based on pre-registered questions. The three researchers gave a more realistic and appropriate answer based on their experience at Minds Lab. They shared their experience at the company and in field and communicated with the participants. There were many questions on the study and research of the AI field, such as how non-AI-related majors can work on AI research, what AI trends beginners should follow, and what kind of research papers to read. To these questions, researchers replied that it is important to read papers from various fields, continue to improve one’s personal development, continuously study the field, and work on and enjoy various projects. Other than the pre-registered questions, there were many live questions through chat and real-time. The StartupTing will be uploaded on YouTube, so please watch if you’re interested. Hearing the stories of the three researchers, the future of Minds Lab seems very bright. AI-related services developed by Minds Lab will transcend other companies, because team members continue to work on personal development. Students who are interested and want to experience AI related services including deep learning should consider Minds Lab! We support Minds Lab J Please look forward to the next StartupTing as well. Thank you!

KAIST Startup-ting X Market Kurly Review

Market Kurly, known for its dawn delivery and purple branding, joined Startup KAIST for May’s Startup-ting. Market Kurly is an online commerce company that is well-known to customers. The company has grown exponentially after its establishment in 2015, and is about to be listed on the U.S. stock market. Customers can receive fresh products before 7 am if they order by 11 pm the night before. Market Kurly is the first to launch a fresh product delivery service.

Startup-ting is continuing through Zoom due to COVID-19. If you are interested in the development and rapid growth story of Market Kurly, pay attention to the review of Market Kurly’s Startup-ting!

1. Market Kurly’s ultrafast growth: a realistic service development story that increased daily sales from 500 million to 6.7 billion won – CTO Sangseok Lim

Beginning with an introduction of Market Kurly’s services, CTO Lim also gave a lecture on the logistics service. CTO Sangseok Lim joined Market Kury at the series C stage to manage and systemize the organization operation. (The lecture below is described in the perspective of the speaker).

Market Kurly is known for its dawn delivery mainly in the capital areas. Dawn delivery is a service that delivers by 7 am if users order before 11 pm the night before. Now, we are expanding dawn delivery to Chungcheongnam-do regions such as Daejeon and Sejong, and looking for ways to expand to areas like Ulsan. After Jeon Ji-hyun’s TV commercial, Market Kurly’s reputation has increased. The CEO of Market Kurly checks products meticulously every Friday to pursue high quality products and create user friendly contents. We provide all information necessary for customers to understand the products in detail, and before COVID-19, we would even go to the United States to take photos of products.

Market Kurly is working hard for a unique branding through innovating packaging. We provide the freshest delivery compared to Rocket fresh and Oasis. Our goal is to deliver Hagen Daz ice cream with a hard surface, the tail of mackerel still frozen, and abalones as fresh to the customers. In May, dawn delivery boxes will be changed from paper boxes to Kurly’s purple box (recyclable packaging).

Market Kurly experienced a rapid business growth as well as a rapid traffic growth. There is still a 5% chance of growing, and the basket size (the amount of items in an order) will become important. Market Kurly is currently in the series E funding stage and preparing for Pre-IPO. Even in the competitive online fresh food market, Kurly is maintaining its lead. Kurly is transforming from a functional to an agile organizational structure, increasing developers from 20 to 200, and introducing an audit-level service development process. The daily orders have reached 120,000 nationwide, and we are considering expanding categories and upgrade into a Kurly Market Place. In 2018, we transferred from IDC to AWS while maintaining out monolithic service structure to take advantage of the cloud.

The rapid growth of Kurly is greatly due to the COVID-19 outbreak. In parallel to the rapid business growth, the traffic volume and volatility have amplified. We experienced the limitations of a monolithic structure, replacing and upgrading to MSA. We have also achieved efficient handling of traffic and operational reliability through expanding AWS Cloud on-demand EC2 and RDS. On the topic of building and evolving the Kurly commerce development team, it is showing a cocktail effect, and both JAVA and PHP exist. It is necessary to understand and use various languages and frameworks to know the limitations, and consider costs for recovery. When there is a large scale traffic, CQRS (command and query responsibility segregation) is required. After the introduction of a DynamoDB and ElasticCache based CQRS, the traffic has changed. Lastly, although there is a shortage of developers, we are working on the problem of growing the basket size.

In the recommendation section, Kurly used to mix curation-based and rule-based automation methods, but has introduced an AWS Personalize system (a personalized recommendation system at a minimum cost in a short period of time). After developing the system and deploying A/B tests, the system showed an improvement in shopping cart conversion rate to 358%, but at the cost of a decrease in the size of the shopping cart by 0.002%. Originally, the customer center tasks were done manually through excel, but is now being optimized through AI/ML based task optimization and order automation.

Amid the growing number of automated distribution center technologies, Markey Kurly uses a PDA picking system, QPS (Quick Picking System), and Cross Belt Sorters to handle the large-scale orders. However, robot systems have not been fully introduced. We still need logistics processing technology and logistics center SW technologies to handle large-scale orders. We are currently developing various algorithms, and methods to minimize picking when stacking products.

Data Analysis Team (Advanced Analytics Team) – Sangrae Noe

I am currently working at the Advanced Analytics Team at Market Kurly and joined the company since September 2016. I also conduct algorithms, analytical data, and decision making on data trainings. I will explain about the teams related to data analysis, and how the data teams contribute to the business. Kurly is internalizing its capabilities to provide the best service, and data analysis takes an important role in this. The amount of data and the complexity to solve issues are increasing. The Advanced Analytics team predicts data Ops, data policies, data & ML infrastructure in the data front end, and also designs dashboards.

I will also explain about the trial and errors over the 5 years of Kurly’s data. In 2015, we used to work with excel. The work was fast, but the system was weak. Due to security and safety issues, we connected the business and data using AWS since 2016. We created a system to collect the necessary information and solve problems in a timely manner. Since then, the era of Data product has opened, and we have been solving advanced problems using the M&L Dashboard. We are now working with a data governance system. For data policies, the information needed for each team design and analytics tasks are required. We develop and analyze data infrastructure, design systems for data analysis, and lastly, develop infrastructure systems for Kurly services.

Q&A and Discussion

Since there were many customers using Market Kurly’s service, there were more questions than any other startup-ting. Besides the pre-registered questions, there were many live questions in chat, making the Q&A session very lively. The session was followed by a variety of topics, including development, Kurly’s service, and changing jobs (Please refer to the video for more questions).

CTO Sangseok Lim replied to various questions during the Q&A session. He added that the review board is being redeveloped completely and that Markey Kurly is the best in PHP development, continuing to challenge and develop new areas.

Due to COVID-19, Market Kurly had a surge in customers. Today’s Startup-ting was a time to learn about how Market Kurly has developed since 2015. It is amazing how fresh products are delivered by dawn, and the meticulous system that allows it. Kurly will continue to develop its service and expand nationwide to deliver fresh products. We hope Markey Kurly challenges new areas and grow further. Please look forward to the next Startup-ting!

KAIST StartupTing X Upstage Review

Today’s StartupTing is joined by Upstage, an AI startup created by AI core developers from Naver, Kakao, Nvidia, and Google. CEO Sung Kim is known for his deep learning lecture, and is currently a professor at the Hong Kong University of Science and Technology(HKUST). If you want to learn about AI and receive mentoring, stay focus on Upstage’s StartupTing.

Career Building in the AI era – CEO Sung Kim

CEO Kim started the StartupTing with a lecture on career building. The lecture will be described from a first person perspective below.

The domestic industry has changed from physical, digital to AI Beneficial. I thought about how to AI transform the myriad of problems in the industry, and I felt the need to create value for a new field of AI modeling, engineering -> cost reduction. I felt that AI planners were necessary to create this value, so I started Upstage. I wanted to create a self-improving AI (SAI) platform, so I named the company Upstage.

We create AI models from a base model that recognizes data, and are planning to make a platform as well. Upstage is a startup founded by AI developers and business leaders with experience in AI services. We also look for talented AI members that will join us in our AI platform.

A story never told of school, company, and startup – CTO Hwalsuk Lee

Before CTO Lee started on his AI lecture, he told his career story as a KAIST senior.

[Career story of CTO Hwalsuk Lee]

After obtaining a Ph.D. on video codec at KAIST, I worked at Samsung Techwin for 15 years. While working on CCTV technology, I read a Deep Mind paper and felt the urge to pursue AI, so I bought GPU and started research. To focus on AI, I moved to NCsoft to develop a bot for one-to-one battles, researched generative models for game development efficiency, and created lectures on autoencoders. In particular, the git repository consisting of generative models using Tensorflow, attracted attention, leading to a collaboration with Google and Reddit.

In order to provide AI services to more customers, I moved to Naver and advanced the OCR (Clova OCR) field. Due to this, I wrote 11 papers and also won the OCR World Championships. After being recognized the contributions and developing dozens of domestic and foreign services, I became responsible for the overall computer vision, including OCR, and joined CEO Kim to start a business.

[School vs Company: A perspective from AI development]

In school or research, one usually looks for a better model for a fixed dataset or evaluation. Research focuses on developing AI models that draws new technologies accordingly with a fixed number of test sets, test sets, and evaluation methods. However, in AI development, there is no fixed train set, test set, and evaluation method. Thus, there may be a gap between a quantitative assessment in the development environment (offline test) and in the real world (online test). Since quality of the service is important, it is critical to design offline tests similar to online tests, and derive specifications of the model.

[Company – AI Composition of organization]

You need members to prepare and manage the quality of the aforementioned data. The company needs a workforce that develops tools for efficient data and modeling, and people who manage the overall quality of models, like data curators, modelers, and IDE developers. If the tech team also serves AI models, more workers are required because there is more work to serve the model to end devices.

One must ensure that data accumulates and consider service requirements when developing services. Thus, the first thing to do when developing an AI model to a service is preparing training data sets. For example, one must define the type, quantity, and target data.

e.g. AI development that predicts Latex expressions when taking pictures of formulas that are difficult to input

To prepare a training dataset, an AI model must be designed, and to design an AI model, a training dataset is needed. It is important to converge as we repeat this process, and now there are a lot of data manufacturing companies that we outsource data production.

[Company vs Startup]

Startups go through the process of Scope Project-> Collect Data-> Train Model->Deploy in Production.

Startups nowadays have big data like any other companies. In the Collect data part, we have to see if data accumulates and check the quality of the data.

In the Train Model part, there are tools that efficiently use GPU, government funding, and startup support businesses that startups can utilize. However, the most important part is the Scope Project part. It is important to develop an AI model according to the service goal. The content below is a worthy consideration when selecting a company depending on the two types of businesses.

Business Centric AI: companies that have grown already – the business before AI was incorporated, existing modules, business flow do not change much

AI Centric Business: startups – place AI at the center and define a new flow (AI Transformation).

e.g. traditional car manufacturers vs Tesla -> or completely new services/businesses.

To summarize, companies like large corporations already have a set rule on AI, so it is difficult to change the stage considering the service stability. On the other hand, startups have no limit, so it is relatively free to create a new stage to introduce new technologies.

AI research and development at Upstage – AI researcher Sungjoon Park

AI researcher Sungjoon Park told in-depth stories about AI research and development at Upstage, focusing on how to process data and NLP. Referring to the StartupTing video will be useful to learn more about AI.

In 2018 Glue, in 2020 BART, and this year GEM appeared. Our team participated in the KLUE project (Korean Language Understanding Evaluation), with 31 co-researchers, 11 collaborators, and 10 sponsors.

AI developers should also consider ethical issues on data sets. It is important to think about how to deal with these issues from a model evaluation perspective. Recently, there are many competitions to elect AI developers and more AI experts are being trained. There was also a test on domain adaptation by Riiid.

Q&A and Discussion

CEO Sung Kim, CTO Hwalsuk Lee, and researcher Sungjoon Park each answered pre-registered questions based on the categories of AI career, AI development, and AI technology.

Looking over a few questions, to the question “Which is better – a company with excellent technology or a company that analyzes data well?, they answered that “a technology company can learn from people around them, and a data analytics company can try many things with the abundance of data, but to choose between them, we prefer a company with excellent technology”. To the question asking what non-AI majors should study to enter the AI industry, they answered that it is important to build a strong foundation with basic computer science courses. To the question asking what ethnical guidelines exist and which ethical value they consider most important, they answered that there are a lot of guidelines on research in universities and companies to refer to.

Besides pre-registered questions, there were many more questions asked live. In the StartupTing of Upstage, there were about 200 participants. There was even a reporter interested in startups that joined to write an article. The interest in AI is increasing as the AI era accelerates. We hope more AI experts appear in KAIST that will lead to even more startups.

Please look forward to and stay tuned for the next StartupTing!

KAIST StartupTing X Bagelcode Review

March’s StartupTing is joined by Bagelcode, a global gaming company with 40 million users from 200 countries. Founded in 2012, Bagelcode has studios in Seoul, London, and Tel Aviv, and secured Series D funding. Let the StartupTing begin!

Bagelcode’s StartupTing begins with a lecture by CEO Kay Yoon and is divided into three parts, including stories from KAIST seniors. The lectures are described below from a first-person perspective.

1. The Startup Story of Bagelcode & Becoming global is no longer an option, but a necessity – CEO Kay Yoon

Bagelcode is a global company with 5 studios in 5 countries. Influenced by seniors at KAIST, I always thought about starting a company. While working at a company in my early 30s, I decided to start a business. It’s been a long time since I left the app development field, so I started by gathering developers. At first, I didn’t know what business to do, so I went through 5 pivoting (webtoon, web-service, social dining, fashion, game) after trial and error. At that time, KakaoTalk and Anipang were trending items, so I developed a KakaoTalk game called Battle Pangpang, but did not achieve commercial success.

The first time I entered the global market was after my failure of the game, and I decided to enter the U.S. market. I thought about casino games that can only be played in the U.S., and began cold emailing companies. As a result, I launched the first social casino game with Gamevil USA. After analyzing American companies and studying how to make a business successful, I worked with Big fish games and attracted global publishing. I adjusted my workstyle to Western culture by introducing calendar invitations and video meetings. Working three years in the U.S. made me, I felt the need to continuously publish and start my own business. Since then, we received Series B funding and established Bagelcode USA. The U.S. branch actively recruited foreigners with experience in the industry to Seoul. Coincidentally, I acquired a company in the U.K. and established Bagelcode UK. After the UK M&A, I recruited experienced members from Israel and established branches in Israel, Ukraine, and that is how Bagelcode became a global company.

2. From a No. 1 Global healthcare Silicon Valley startup to Bagelcode DATA & AI Director – AI Leader Joohyun Kim AI Leader

My name is Joohyun Kim, and I was previously a developer at eBay and am currently the AI leader at Bagelcode. When I choose a career, I consider growth, challenge, and fun the most important. I met CEO Yoon while working as a military industrial agent, and studied machine learning and data while pursuing my Ph.D. in the States. While working at the Trust Science team at eBay, I gained hands-on experience with data. Since then, I joined a healthcare startup MyFitnessPal. The company was a popular startup that later became acquired by Under Armour in 2015. The challenge of becoming a leader at a data team at a data-oriented company and the potential of Bagelcode drived me to join Bagelcode.

Eventually, I made the decision based on what I want to do, what career I will do in the future, and what story I can tell. I am now working at Bagelcode because of data-AI machine learning, startup exit, growth, challenge, and my team. I am in charge of the technical and strategical parts of data and AI. The data and AI team main tasks are Data platform/Data engineering/Data science. Bagelcode is not just a game company, but a company with smart team members, IT technology, fast growth, and well-developed communication. Bagelcode also values data by monitoring the data dashboard.

The two important things in data is cumulative number of users and the game action count. Bagelcode has secured more than 50 million users since the 2019 acquisition. The game action count (actions of users in games) has increased since 2017, and Bagelcode has accumulated data worth 60 billion KRW. Data is managed by over 600 dashboards and 1700 event tables in 5 games, and 7TB of data is processed daily. These data is used in AI and machine learning to strategically predict what each user will do at a certain point. I believe it is necessary to automate monitoring of key indicators in games. Bagelcode is a fast-growing company backed with data and technical specifications unlike other game companies.

3. Establishing and operating a development team in Korea, UK, and Israel after the global M&A, Head of EXP Donghoon Cha

I joined Bagelcode in 2012, and am currently the head of EXP. I worked as a software engineer for 5 years, covering client and back-end areas. I also served as an all-rounder working on various teams as a data analyst for 6 months, product owner for 2 years. I will introduce the most global team at Bagelcode, JPJ. After the M&A, the team divided into three regions (Seoul, London, Israel) and developers also worked in Canada, New Zealand, and Portugal. Due to this division, it was difficult to manage the team. It was imperative to create a collaboration protocol, so I went on a business trip to London and had a marathon meeting all day. We had daily meetings to improve the work pace and collaboration process. We didn’t conduct the regular scrums and there were also problems due to culture and language difference. However, we managed to integrate people of different nationalities into one field.

As a global company, time, distance, language, and culture problems are inevitable. Because we are one team in different parts of the world, business trips were frequent. However, due to COVID-19, this became even more difficult so we tried to solve problems through online meetings. We communicate using English, but sometimes we miss details and experience difficulty in communication because of different nationalities. Because we are a global team, all members must work hard to mitigate these differences. Nevertheless, the challenge and joy of working at a global team is great.

Q&A and Discussion

In the Q&A session, many students participated and submitted questions in advance. Questions were divided into fields and answered.

1) Startup

To the question regarding the challenges and timing of startups, it is important to meet many people and find mentors who are experienced with startups, because problem-solving skills are important in starting a business.

2) Global capabilities

Regarding questions on how Bagelcode became global and the difficulties becoming global, the Bagelcode team gave a simple example adding on to the lecture earlier. In addition, the global recruiting problem is inevitable, so the team will continuously work on this.

3) Military service exemption, etc.

Bagelcode recruits 2-3 people every year for special military service exemption. In the case of interns, they do not differ much from regular employees, but tend to work for about 6 months. We look for members who are flexible in communication and relationships. We select people who fits these descriptions: “working with this person is fun”, “the best welfare is talented individuals”.

Bagelcode is currently recruiting members in all fields, so visit the homepage (https://bagelcode.recruiter.co.kr/appsite/company/index) if you’re interested.

There were many live questions as well, and the team members answered them sincerely. Bagelcode was the most global company that joined us for StartupTing. There are branches in England, Israel, Canada, and New Zealand, and we are proud that the center of Bagelcode is Korea. We expect greater growth for Bagelcode, a global company that focuses more on data than games.

Please look forward to the next StartupTing!

2021 GRAFFITI Startup Festival – Report

GRAFFITI Startup Festival 2021

ICISTS, which consists of KAIST undergraduates, have been hosting ICISTS, Asia’s largest international university conference, for the past 16 years. In 2019, ICISTS hosted “ICISTS 2019” as well as “1st GRAFITI” to spread the startup culture.

The event was held to form bonds and maximize capabilities of passionate and spirited university students interested in the startup ecosystem.

“GRAFFITI 2021” is a non-face-to-face startup festival that takes into account the COVID-19 situation and further develops “ICISTS 2019”.

The conference aims to provide the experience of initial startup process, and expand as well as increase the inflow of the startup ecosystem.

WHY GRAFFITI Startup Festival?

ICISTS is very similar to a startup.

ICISITS pursues a greater growth through ‘meeting people’, and GRAFFITI was a new challenge and ambition for the organization.

Through GRAFFITI 2021, the organization aimed to spread the vision and value of ICISTS.



The investment game proceeded in the form of a round survival game, and each round had a different concept.

An example of concepts was “How the current team was formed & the team culture”. Like so, the investment game focused on bringing out the values and philosophies of student startup teams rather than their technologies and ideas.

In the middle of the session, YouTuber “EO” and CEO Junghee Ryu of Futureplay joined the panel and commented on the stories of the student startup teams, establishing an open conversation.

The final round consisted of 3 teams who did live pitching, and winners received awards and prizes.

1st place was ‘labNote’, a team that develops a data record system and machine learning solution for research,

2nd place was ‘Waddle’ who runs a machine learning-based mobile shopping application ‘Sori Market’ for people who are visually impaired,

and 3rd place was ‘Maring’, a company creating depression-relieving contents.

Participants who invested to teams were also rewarded depending on the cumulative return on investment.

Keynote Speech & Talk Concert

CEOs of active startups in the ecosystem joined the festival and shared their startup stories and challenge spirits.

CEO Yuwon Lee of ‘Banjiha Games’ and CEO Chaerin Lee of ‘CLASSUM’ joined, and Youtube Live streaming made the talk concert much more lively.

Under the hosting of the YouTube creator “EO”, CEO Sedong Nam of ‘VoyagerX’ and CEO Jay Choi of ‘Button Soup’, CEO Sungjun Jung of ‘Mr.Mention’ discussed topics on startup culture of 20s and student startups. It was a time to talk about various subjects, and through Youtube Live, the speakers received various comments and questions.

Team Project

Team project was designed for participants to experience the process of starting a business.

Each of the nine startups participating in GRAFFITI 2021 was matched with 3 or 4 teams.

The key of the team project is for participants to think like a member of the startup and experience the shortened startup process with the matched startup.

Step 1. Team matching – matching with startups

Step 2. Get to know pt. 1 – startups introduce their companies / participants identify and understand matched startups

Step 3. Get to know p3t. 2 – analysis and ideation of startups

Step 4. Verification – verify marketability and feasibility of ideas through market research

Step 5. Design – develop ideas through prototyping and creating business models

Step 6. Presentation – present in front of other participants and startups

In the team project session, there was a lecture and Q&A session by Mario Garcia Lee, CEO of Lighting up Ventures to facilitate the activities of the participants.

During the final evaluation, participants received comments from the leading domestic VCs such as ‘Mirae Holdings’, ‘FuturePlay’, and ‘Mira Partners’.


Most people are afraid of failure, and they fear starting a business due to the potential loss and failure.

Fail-con was like a radio station and delivered a message that failure was not an end in itself, but a nourishment for greater growth.

Startups that experienced many failures, VCs who observed these failures, and participants discussed and tried to see failures from a new perspective.

Fail-con consisted of VC talk, speaker pitching, and failure story interviews.

During the VC talk, CEO Shinhyuk Kang of ‘Coolidge Corner Investment’ and CEO Mira Park of ‘Mira Partners’ talked about the failure of startups from an investor’s perspective.

During speaker pitching, the stories of 3 speakers who experienced failures from startups and used the failure as an opportunity to live fuller lives provided a new perspective on failures.

In the failure story interview, two participants shared failure on a deeper level and it was a time to allay fear on the word “failure”.

KAIST StartupTing X MINDs Lab Review

2021’s first StartupTing is joined by a renowned AI company, MINDs Lab. MINDs Lab is the best and only AI expert organization equipped with AI algorithms to AI applications and services. The company has recently launched a subscription AI platform ‘maum’ open to anyone in the cloud field.

In today’s StartupTing, CEO Hong-Seop Choi introduced MINDs Lab and researchers from KAIST explained the characteristics of various AI fields. About 80 students actively participated in StartupTing through Zoom.

MINDs Lab Introduction – CEO Hong-Seop Choi

MINDs Lab wishes to create a world where anyone can use AI and become a company that can affect the world with AI. The company is the first Korean company to partner with amii, one of Canada’s top three AI research institutes, and generating single sales in the AI field. In addition, MINDs Lab is developing AI programs and algorithms to provide a cloud service using an AI engine, and adding edge computing to its AI platform ‘maum’.

When CEO Choi opened the maum AI cloud platform, he received a lot of collaboration requests not only from large companies but also small and medium-sized startups. He mentioned that all startups are related to AI somehow, and that MINDs Lab started its business in the AI field early. Using the ecoMINDs program, any startup can utilize AI easily. The company is currently establishing the 2nd Center for ecoMINDs in Daejeon.

After introducing MINDs Lab, CEO Choi explained the benefits of joining the company. MINDs Lab headquarters is located in Pangyo, and it is possible to work either in Pangyo or Daejeon, as well as at home due to COVID-19. In addition, MINDs Lab has A-backs. The company is looking for globally talented people, who can also work at the Silicon Valley Branch and in Toronto Canada. Applicants can apply for Full time/internship/part-time (only Daejeon office)/industrial functional agent, professional functional agent, etc., and there will be interviews and coding tests.

Brain Audio – Researcher JunHyeok Lee

Senior researcher JunHyeok Lee gave a friendly and easy explanation of what the Brain Audio team of MINDs Lab does. Researcher JunHyeok Lee is a graduate of KAIST Mechanical Engineering currently working at MINDs Lab.

The Brain-audio team is working on the development of audio-voice signals, and servicing a streamer TTS (Text to speech). MINDs Lab’s TTS service converts foreign language text into speech data and successfully created a voice conversion system with Cotatron. Research Lee showed real-life examples, increasing interest in participants. MINDs Lab is applying speech recognition in government projects and actively researching speech to text. The company’s voice filter field is also recognized by Google for its technicality.

MINDs Lab Brain Vision Team – Researcher Hyoungkyu Song

StartupTing was followed by a lecture by Senior Researcher Hyoungkyu Song from the Brain Vision Team. Researcher Hyoungkyu Song is currently a masters student at the Department of Bio and Brain Engineering at KAIST and is working as an industrial agent.

Researcher Song said that courses at school, and club activities helped him grow. He mentioned that to proceed in the field of deep learning, it is important to have experience in projects. Deep learning requires a lot of supplies, and MINDs Lab has a lot of GPUs. At the company, he is coding the contents in research papers. Looking at the direction of AI scientists compared to developers, it is important to set metrics, experiment to improve performance, and preprocess data. Lastly, it is important to constantly research, discuss and develop. Currently, the Brain team consists of people from various departments and is also doing code review. There is also a vc in data sets.

As there are more related academic conferences, you can do research on what you want to do at MINDs Lab. The company is currently focusing on the Lip sync-seen model, which creates images that seem to realistically speak for a particular person, and continuously developing to create optimized models through various studies, such as Holistic-3D Reconstruction, FaceShifter, and Point cloud completion. The company is also building models using data from customers, such as a tooth model that locates a suitable place to add braces and a photoshoot model that recommends clothes.

The brain team of MINDs Lab is composed of young people with strong potential, and you can research and directly experience what you have learned in courses at KAIST.

Q&A and Discussion

During the Q&A session, we received a lot of pre-registered questions and also real-time questions through Zoom. The CEO and researchers answered many questions about working at MINDs Lab, the skill sets they see in applicants, etc. They answered that positions are divided into development and research positions and subdivided according to the roles. The service positions do not require any research experience. Coding tests see the ability to review research papers and analyze models. To the question about which field developed the most in AI, they answered that MINDs Lab is ahead in the contact center-voice recognition chatbot and AI human voicebot, and that the CCTV vision traffic-related abnormal behavior control will be commercialized in the future. MINDs Lab is currently not doing direct B2C business and have no plans to do so. There were many more questions during the session.

MINDs Lab is also working with the bio field to predict new drug candidate materials. MINDs Lab has 16 AI-related patents, and wishes to create an AI engine that anyone can use. Anyone can join MINDs Lab and do research on what they wish to. If you want to grow and become an AI scientist, take a good look at MINDs Lab. Please look forward to the next StartupTing 🙂

2020 X-IST Startups Competition – Report

Five major institutes of science and technology (KAIST, GIST, DGIST, UNIST, POSTECH) hold annual joint startup competitions among promising startups from each university to strength the solidarity of startups and propel competition. This year was the 7th competition, and the competition has produced many startup teams so far. Let’s take a look at the 2020 X-IST Startups Competition!

Preliminaries in each university

Each university selected 2 promising startup teams to participate in the X-IST competition final. This year’s preliminary was conducted online due to COVID-19. KAIST selected ‘Aldaver’ and ‘Spidercore’, finalist teams from a lab research-based startup program E*5 LabStartup.

Main event (Final round)

On December 4, 10 teams selected through preliminaries from each university competed against each other through a mock investment pitching method. The presenting teams and evaluation committee were present at the site, and there were about 200 participants watching online. Each team had 10 minutes for the presentation, which was followed by a 10-minute Q&A session. 5 external experts (investors and startup experts) recommended by each university evaluated the business feasibility and excellence of participating teams.


After the presentation, the results of the competition were tallied and there was an award ceremony at the site. Team ‘Aldaver’ from KAIST presented about ‘Developing the next generation artificial cadaver technology using functional hydrogel’and won the grand prize. Aldaver received a prize of 5 million won and an award from the Ministry of Science and ICT. 2 teams received the awards of excellence, and 2 teams received the awards of encouragement. There were teams who didn’t receive awards, but teams learned a lot from other teams and recognized them as competitors in good faith.

Which teams will be competing in the future? We look forward to next year!

17th E*5 KAIST (`20 Fall) – Online

KAIST Student Startup Audition Program

E*5 KAIST (17th)

  • 18 Teams Participating

    Total of 64 Participants

  • Duration

    2020. 9. 22 – 2020. 12. 4

  • Awards

    1 Grand Prize / 2 Awards of Excellence / 1 Special Prize / 1 Activity Award

Grand Prize


Award of Excellence (2nd Prize)


Award of Excellence (3rd Prize)


Special Prize


Activity Award


  • Mentors

    17th Mentor Group

Yonggun Kim

Yonggun Kim


Bluepoint Partners VP

Hwaseong Jeon

Hwaseong Jeon



Junghee Ryu

Junghee Ryu


FuturePlay CEO

Ki-jun Kim

Ki-jun Kim


Kakao Ventures VP

Tae-Yeon Jeon

Tae-Yeon Jeon


BonAngels Venture Partners General Partner

  • Seniors

    E*5 KAIST Graduate Founders

Jeungmin Oh (3rd E*5)

Jeungmin Oh (3rd E*5)

HayanMind Inc. CEO

Chang Gi Hong (8th E*5)

Chang Gi Hong (8th E*5)

H2K Inc. CEO

Howook Shin (13th E*5)

Howook Shin (13th E*5)

SelectStar Inc. CEO

  • Details

    MIssion + Activity

Out of 25 applicant teams, 18 project teams were selected through the first mission. The selected teams then developed their ideas through each mission, which consisted of training, coaching, and evaluation from the mentors. Due to COVID-19, the entire process of the 17th E*5 KAIST program was held online. During the Final Mission Evaluation, 11 teams made their final pitch, and were evaluated by mentors.

Mission Training

Mission Coaching (By Seniors)

Mission Evaluation

1 Mentor was assigned to each team for training, and 3 Seniors were assigned to coach the teams on a regular basis. After the training and coaching stages, teams were selected through evaluation to receive additional funding.





Field Trip


During the program, teams visited mentors (Mentoring), met with field experts to solve issues (Networking), visited startups (Field Trip), conducted market research (Research), and collaborated through activities and team meetings (Teamwork). All activities were uploaded and recorded on ‘CLASSUM’.

Build your Startup,


2020 KAIST Startup Awards – Report

In addition to programs that foster entrepreneurship and spread startup culture, KAIST has supported startup audition programs. Many startup teams were produced in 2020, and ‘2020 KAIST Startup Awards’ was held to select the best startup teams. Due to COVID-19, Startup KAIST could not provide the opportunity to attend CES. However, through the awards Startup KAIST discovered startup teams and plan to foster them as a unicorn company through continuous support.

Selection Process & Procedures

2020 KAIST Startup Awards was joined by 10 senior entrepreneurs, investors, and related specialists as evaluation committee members. The committee evaluated the 1st preliminary application round focusing on creativity, specificity, and business feasibility, and selected 11 teams. In the 2nd presentation round, the committee evaluated teams based on their startup passion, development and business capabilities, business feasibility, and excellence in commercialization plans, and selected 4 final teams.

Final Teams

At the 2020 KAIST Startup Awards, a total of four excellent teams were selected, including 1 Grand Prize, 1 Award of Excellence (2nd Prize), and 2 Awards of Excellence (3rd Prize). The Grand Prize winner was ‘Chequer’, a startup developing a data access control and audit trail SQL collaboration solution. 2nd Prize was given to SelectStar, and 3rd Prize was given to ‘QuantumCat’ and ‘Dolbomdream’. Congratulations to the final winners and we look forward to the success and future of the teams!

KAIST Startup Ting X Curocell Review

2020’s last Startup Ting was with Curocell, a company developing innovative anti-cancer immune cell therapy, CAR-T. Curocell is a company co-established by Professor Chanhyuk Kim of the Department of Chemical & Biomolecular Engineering at KAIST. The Startup Ting was originally planned as a company visit, but due to COVID-19, it was held online to introduce the company and its technologies, and demonstrate online lab tours.

Introduction to Curocell & Anti-Cancer Immune Cell Therapy – CEO Gunsoo Kim

CEO Gunsoo Kim explained about the history of anti-cancer drugs, and the anti-cancer immune cell therapy under research at Curocell.

Until 2011 there were only anti-cancer drugs, but anti-cancer immune antibodies appeared and the methods of chemotherapy have changed. Previous anticancer drugs worked on a mechanism that directly targets cancer cells, but anticancer immune cells strengthen the immune cells in the body and remove cancer cells through immunity. Anticancer immunotherapy can overcome the limitations of traditional anticancer drugs, and is a very important treatment that will become the lead in the anticancer drug market,

In August 2017, Curocell received the first approval of the FDA for CAR-T anticancer immune cell therapy. The CAR-T therapy separates the immune cells of the patient, manipulates genes, and goes through mass cultivation, and is injected back to the patient. It is the most innovative anticancer drug developed so fat, and is expected to cure terminal patients with one dose. However, the anticancer drug market is at the initial stage, with only two products approved as recent anticancer drugs worldwide. Curocell’s technology will be able to treat terminal patients, and can be checked whether cancer treatment has effectively worked within 30 to 60 days with a single dose. The CAR-T treatment clinical trials are actively being carried out mainly in the U.S. and China, but not in Korea due to the lack of technology and experience. Curocell is preparing to receive a license for clinical tests next year.

CAR-T cell therapy is a new paradigm of chemotherapy, which genetically manipulates a patient’s T cells separated from the cancer cells and inserted back to the body. Curocell manufactures T cells separated from cancer patients by recognizing cancer and self-activating CAR genes. The CAR-T cell therapy is a competitive field amongst global pharmaceutical companies.

Curocell is the corporation developing CAR-T cell therapy which is world-widely taking center stage and has received a cumulative investment of 61.5 billion won as of 2020 since its establishment in 2017. Curocell was founded by collaboration with CEO Gunsoo Kim who has long experience of pharmaceutical company, Prof. Chanhyuk Kim of KAIST and Prof. Hyunbo Sim of Ehwa Womans University.

[About Curocell’s OVIS™ CAR-T technology]

Curocell is developing a new OVIS™ CAR-T technology that combines CAR-T technology and immune suppression technology.

OVercome Immune Suppression

The next generation CD19 CAR-T, which is featured with the 2nd generation OVIS™ technology, is expected to treat patients more effectively. The company consists of 40 product development-oriented employees who have extensive experience in various fields, such as R&D, GMP production, and clinical development.

The most important thing is to cooperate with hospitals, and Curocell has a strategic partnership with Samsung Medical Center. Establishing a GMP facility in the hospital, Curocell is the world’s first to do a clinical launch of the next generation CAR-T technology that can treat blood cancer as well as solid cancer.

->Here is the competitive advantage that Curocell has:

  • Immune suppression CAR-T technology & more: globally leading the OVIS™ CAR-T technology
  • Korea’s first clinical launch of the next generation CD19 CAR-T: Next generation CAR-T R&D, GMP production, etc.
  • CD22, BCMA, EGFRvlll, CAR-T & more: capabilities to secure pipelines through the company’s technology and open innovation
  • Cooperation with Samsung Medical Center and GMP facility for CAR-T: Establishing a close partnership with Korea’s top hospitals.

Introduction to Curocell’s Research Institute

CEO Kim additionally explained how R&D is done in a pharmaceutical company and about its research institute. There are three main ideas to consider when developing a drug.

  • Quality by design (QbD, design-based quality enhancement)
  • QTTP (Quality Target Product Profile): how to administer the drug
  • CQA (Critical Quality Attribute): drugs must be specific and consider safety and effectiveness based on previously known facts
  • CMA (Critical Material Attribute)/ CPP(Critical Process Parameter): must consider the properties of drugs, what materials to use, and how to manage the drugs.

E.g.) Breeding cows

  • What kind of cows should we breed? -> QTTP
  • What characteristics of the cow is related to good taste and rich amount of meat? -> CQA: usually based on shipment and delivery standards
  • How should we breed the cows so that they are healthy and have a good balance between weight and fat ->CPP: usually measured in numbers, modified during processing & CMA: difficult to change the material, type during processing

[QbD for CAR-T Development]

  • QTTP: safety, reliability
  • CQA: presence of microbial contamination, Non-T cell content, residual content, amount of CAR-T cells, cell survival rate, cytokine secretion, etc.
  • CMA/CPP: T-cell separation methods, number of cell cleanings, amount of virus used, types of cell culture mediums, duration of cell culture, final amount of cells, freezing conditions (cell concentration, composition, methods), etc.

Curocell’s R&D center mainly consists of 3 teams – discovery team, process development team, and an analysis method development team. The discovery team works with the research team to design CAR cells for new targets, improve OVIS™ technology and perform adaptive evaluation, and perform effectiveness evaluation using an In-vitro/In-vivo model. The process development team establishes CMA/CPP (experiment planning) on new products, sets a control strategy, and improves the production of CAR-T. The analysis and development team develops analysis and validation methods, analyzes samples for CMA/CPP selection, and sets standards and testing methods. These 3 teams lead to QA (Quality Assurance), Production, and QC (Quality Controls) at the GMP facility. Curocell’s R&D center has the following plans:

  • New Blood Cancer OVIS™ CAR-T
  • OVIS™ Platform Improvement
  • Solid Tumor CAR-T
  • Allogeneic CAT-T

Q&A and Discussion

There were many questions about solid tumor and CAR-T, so CEO Kim answered questions in relation to CD-19.

[CD-19 negative relapse]

  • Re-occurrence rate of 20~30% after CD19 CAR-T treatment
  • 50% of the recurrent patients are CD19 negative leukemia

In solid cancer, one of the causes of failure for CAR-T treatments are immune checkpoint receptors. Curocell is trying various technologies based on PD-1 to improve the immune checkpoint receptors and T-cell-CAR-T treatment.

[Questions about startups]

The founders of Curocell did not invest their own money, but came so far by receiving series A, B, C by appealing the experience and capabilities of the founders and preparing clinical tests. To the question “How should we prepare our career to develop drugs?”, CEO Kim answered that at Curocell there are no graduates from colleges of pharmacy and there are a lot of cases where employees learn while working at the company. Curocell’s Q&A session has a lot of information related to developing new drugs, so please refer to the Youtube video for more details.

Lastly, Curocell showed a video tour of the R&D center. Although the tour was online, it was a time to check out Curocell’s research center and see how the OVIS™ CAR-T was being developed.

During Curocell’s Startup Ting, we realized that it is not long before we conquer cancer! We hope the CAR-T treatment is actively researched and developed to create a new drug for cancer patients. Although there are difficulties of the drug regarding solid cancer, a new drug will be developed by cooperating with hospitals. We hope Curocell receives clinical approval in 2021 for the CAR-T treatment, and shine in the global market!

During Curocell’s Startup Ting, we realized that it is not long before we conquer cancer! We hope the CAR-T treatment is actively researched and developed to create a new drug for cancer patients. Although there are difficulties of the drug regarding solid cancer, a new drug will be developed by cooperating with hospitals. We hope Curocell receives clinical approval in 2021 for the CAR-T treatment, and shine in the global market!