I spent three months interning with the Physical AI Team at KRAFTON AI, and it turned out to be a truly eye-opening experience. Every morning when I arrived at the office, there were always lively conversations happening at the whiteboard, and it wasn’t unusual to see teams from different areas exchanging ideas freely. That kind of open and collaborative culture is what stood out to me the most at KRAFTON AI.

Even though each team was working on very different challenges, people were quick to share advice and knowledge, all connected by the common thread of AI. That spirit of openness was not only inspiring, but also a big help for the relatively new Physical AI team as it built its foundation.

My work focused on bringing robotic movements into the real world, which was a refreshing change from my previous lab experience that was mostly about academic, tightly controlled experiments. Since the gap between simulation and reality is such a big hurdle in robotics, this gave me a chance to deepen my interest in that challenge.

Although my internship lasted only three months before I returned to the lab, I left with the invaluable experience of exploring new directions alongside amazing colleagues. I’m really excited to see what lies ahead for KRAFTON AI—and I’ll be cheering from the sidelines.

During my internship at KRAFTON’s Deep Learning Division, AI Strategy & Operations Team, I had the opportunity to contribute to strategic initiatives. I worked on various research projects and interviews, covering AI research trends in the gaming industry and among major competitors, which informed a long-term strategy project for the division. Later, I also participated in the mid- to long-term roadmap project for production innovation, allowing me to gain broader experience.

The team played a pivotal role in aligning AI research and development with the strategic direction of the division, and I found it especially meaningful to be part of the discussions on the future of KRAFTON AI. Above all, I am grateful to the team leader and colleagues who encouraged me to engage deeply in various discussions and projects, which enabled me to learn tremendously. Thanks to the team’s open and constructive feedback culture, collaboration was seamless, and I truly enjoyed working in such a positive environment with outstanding teammates.

I was also inspired by the division’s brilliant researchers and the passionate yet welcoming atmosphere they fostered. Observing new agendas and projects constantly emerging during my internship, I often wished I could have contributed more deeply and meaningfully over a longer period. Nevertheless, it was a truly rewarding experience to reflect on the future direction of KRAFTON AI and to participate in diverse projects that supported this vision.

A company with a promising future, KRAFTON.

This internship turned out to be a short but incredibly rewarding experience.

During my time here, I picked up the habit of reading arXiv every morning like a newspaper, and I learned how to prioritize and tackle tasks in a more structured way. Working on Foundation Model development also gave me a valuable perspective: the work we do today can evolve into downstream applications tomorrow.

Since I’ve now stepped into the relatively small Speech community, I plan to continue on to graduate school and publish research that KRAFTON can actually reference and build on. Compared to a company, grad school comes with fewer time constraints, which will give me the chance to put in even more effort and contribute to creating technologies that improve people’s lives.

This internship has been a real stepping stone for my growth, and I’m deeply grateful to HS and JW for giving me this opportunity.

My main project was building a dataset collection and preprocessing pipeline for TTS and SLM model development. Through this work, I learned firsthand that no matter how advanced a model architecture may be, “Garbage in, garbage out” still applies. The decisions made at the dataset stage are absolutely critical. To avoid repeating the same mistakes after training, I tested every possible argument combination in the pipeline, listened carefully to the audio outputs, and made sure to honestly judge whether the data truly served our purpose. That process taught me to take ownership and responsibility, even for the smallest details.

Through the KRAFTON AI Internship, I had the meaningful opportunity to apply the concept of Data-centric AI, which I have been researching, to a real gaming environment.

Within the User Modeling AI team, we worked on solving the Anti-Cheat problem in PUBG, specifically detecting users who employed ESP hacks, using both ML and DL techniques. My role was to research data selection methods to enhance the performance of existing Anti-Cheat models. By collaborating with my excellent teammates, sharing ideas, and receiving valuable feedback, I was able to develop a method that achieved robust performance across various scenarios. I was especially delighted that some of the outcomes were actually applied to the live service, allowing me to make a tangible contribution to the team.

At university, I had always wondered how often such data-related challenges would arise in real-world industry settings. During my time at KRAFTON, I was able to experience firsthand the process of identifying and resolving these issues while working with real data, which answered many of my questions.

Although my internship has now come to an end, the passionate atmosphere of KRAFTON AI and the wonderful teammates I had the pleasure of working with will remain in my memory for a long time. I am truly grateful for this valuable research opportunity and experience, and I will continue to sincerely support KRAFTON’s inspiring vision and future endeavors.

During my internship at KRAFTON’s Data-centric AI Research team, I had the invaluable opportunity to explore research topics that could be applied to various games—topics I found deeply fascinating. There were ample opportunities to delve deeply into such fields, and I was particularly impressed by the abundance of computing resources, which allowed me to implement research ideas without major constraints. The environment also encouraged autonomous research, enabling me to stay fully immersed in my work.

Research discussions among team members were vibrant and active. By sharing and discussing each other’s topics, I was able to stay naturally in sync with the latest trends in AI research. Within the Deep Learning Division, there was a well-established culture of actively sharing new papers and interesting AI news. I especially appreciated the strong synergy created by working with highly self-motivated colleagues.

Above all, the greatest strength of the KRAFTON AI Research team lies in its rich portfolio of unique and practical research opportunities grounded in proprietary game IPs such as Battlegrounds and INZOI. This environment offered me a one-of-a-kind research experience that would be difficult to find in other AI organizations, and I am genuinely excited about the future AI research outcomes KRAFTON will present. If given the chance, I would love to return and take part in meaningful and exciting research once again.

I believe choosing this internship before starting my Ph.D. program was one of the best decisions I’ve made. The experience broadened my perspective, enhanced my ability to collaborate, and will undoubtedly serve as a valuable asset in my future research journey.

“Innovation comes from a horizontal and free decision-making structure.” This was one of the most memorable statements I heard from Chairman Chang Byung-gyu during his talks. So when I got the opportunity to intern at KRAFTON AI, I had high expectations—along with a bit of curiosity. "Is such a culture really possible in a 100-person organization, not a startup?"

To my surprise, it absolutely was. Titles were kept minimal, and the organizational structure evolved organically depending on the situation, creating an environment where flexible and autonomous research was truly possible. One thing that stood out was how each team regularly shared updates on their projects with the entire division. This fostered open communication and left little room for silos to form.

I was part of the Data Centric Research team during my internship. Within KRAFTON’s broader research directions—such as CPC—I was given the autonomy to define and pursue my own topic. I could also freely engage with exciting ongoing projects from other teams. Discussions with talented colleagues in areas ranging from LLMs and data to reinforcement learning helped me continuously refine my research.

If you’re looking for an environment where you can work with startup-level autonomy and initiative, while applying AI to the uniquely engaging world of games, I truly believe KRAFTON AI will be a rewarding and enjoyable experience.

My internship at KRAFTON AI was a valuable experience that allowed me to witness the balance between academic research in AI and its practical applications. Rather than staying within the realm of theoretical exploration, I was able to see firsthand how AI is used in actual game development processes and the directions in which it can evolve.

Through sharing research outcomes and engaging in discussions with professionals from diverse fields, I was able to broaden my perspective—not only within my area of expertise but across the AI domain as a whole. Building on these experiences, I had the opportunity to submit a paper to CVPR 2025, one of the top conferences in the field of computer vision.

Additionally, the flat organizational culture, abundant resources, and the supportive atmosphere where everyone actively contributes to each other's growth kept me deeply engaged in my research throughout the internship. Whenever I encountered complex challenges, open discussions with team members helped me find effective solutions. This process enhanced not only my technical skills but also my ability to approach problems from a practical perspective.

I highly recommend the KRAFTON AI internship to anyone interested in game technology or the industrial application of AI. Through this internship, I gained a vivid understanding of how AI technologies are refined into solutions that address real-world problems.