INA Research Group

교수 소개

저는 KAIST 출신이고 여러분들이 하는 개별 연구, 실험 수업 등등 다 경험을 해보았습니다. 수업과 개별 연구를 통해 지금까지 도움이 되는 많은 것들을 배운 경험이 있기 때문에 학생들에게 도움이 되는 환경과 기회를 만들어주고 싶습니다. 연구나 그 외의 일반적인 조언이 필요하다면 언제든지 이메일로 연락 주시기 바랍니다 :-)

당신의 Goal은 무엇입니까?

개별연구, URP, 대학원 진학 하기전에 분명한 목표를 가지기 바랍니다.

석박사 과정

만약 KAIST 대학원 지원을 고민 중이거나 이미 지원한 경우, 연구 관심사에 대해 함께 의논해 볼 수 있으면 좋겠습니다. 저는 학생들이 지도교수를 선택하기 전에 (저를 포함하여) 많은 교수님들과 대화하기를 추천합니다. 여러분이 가지고 있는 연구 관심사에 대해서 하고 싶은 이야기가 있다면 언제든지 이메일로 연락 주시기 바랍니다.

연구 소개

지금 진행하고 있는 여러가지 연구중에 URP 또는 개별 연구 학생이 할만한 주제를 뽑아봤습니다(하단 참조). 몇몇 주제들은 현재 국내외 대학원생들과 진행하고 있는 것이지만, 아직 대학원생들이 배정되지 않아서 학부생들과 새로 시작해볼 수도 있는 주제도 있습니다. 상담 후 자신에 맞는 주제와 배우길 원하는 주제로 정하면 됩니다. 이 외에 항상 아이디어가 있기 때문에 더 많은 주제들도 있고, 학생들 스스로 하고 싶은 주제를 찾아오셔도 됩니다.

학부생 연구의 성공 기준

개별 연구 및 URP의 성공 기준은 본인이 연구에 흥미를 느꼈느냐 못 느꼈느냐라고 생각합니다. 따라서 성공/실패는 본인의 현재 능력과는 절대적인 상관 관계가 없고, 본인의 interest와 동기가 더 중요합니다. 하지만, 학부생들은 본인의 interest를 잘 모를 수 있기 때문에 (저도 그랬고), 그것을 찾아나가는데 도움을 주는 것 또한 개별연구의 목적입니다. 어떤 레벨에 있던지 본인에게 approachable한 task를 주는 것이 교육자의 역할이라고 생각합니다. 저희 실험실은 흥미로운 주제와 자신의 생각을 펼칠 수 있는 환경 그리고 교수와 선배들의 가이드를 제공해 줄 수 있고, 그럴 의무를 가지고 있습니다. 여러분이 이러한 권리를 잘 활용해서 학생으로서 최선을 다하시고 앞으로 나아가 연구자로서 어떤 skill이 필요한가를 배우시기 바랍니다.

🇰🇷 🇺🇸

Graduate-level (Prospective or Current)

I am actively looking for graduate-level students. If you want to apply to KAIST or have applied to KAIST, I can meet with you to talk about our research interests. I encourage students to talk to many potential advisors (including me :-)) before they select one. If you want to talk to me for any reason regarding your research interest, please email me. If you are a KAIST student, I reply to all of your emails; 100% guaranteed.

International Students

If you are not already admitted to KAIST, I cannot do much for you. If you are truly exceptional, I may be able to make your application process smoother. But, most exceptional students do not need my help to get in to KAIST. You can still contact me if you are interested to seek more information. But, sorry; I may not be able to answer all your emails.

Undergraduates

I like to work with undergraduate students in research and provide guidance:

If you are in EE, I encourage you to take EE209, EE205, EE311, EE323, EE324, and other computer science courses such as CS230 (system programming) for your own benefit. You must take EE209 first to take any other courses, but if you have done basic C programming, you can take EE209 and CS230 at the same time. AFAIK, CS230 is a very approachable course. I took it in my freshman year and got an A+. I encourage you to do Individual Research (개별연구) with me after taking at lest 2 or 3 of these courses.
If you are in CS, I encourage you to do Individual Research (개별연구) with me.

Individual Research for Undergraduates (or URP)

I will do my best to provide resources for you to learn about research and make a meaningful progress/contribution in the area of our mutual interest. I spend quite a bit of time with students who are take Individual Research. My goal is to create a fun academic environment through in-depth interactions with KAIST students.

Students Seeking General Advice

Many students contact me for an advice. I like to provide general advice for students regardless of its nature. Please don't hesitate to contact me.

Graduate Students (Masters and Ph.D)

INA research group is actively seeking highly motivated graduate students who are interested in pursuing research in the areas of Internet services and applications, cloud infrastructure, and systems that support artificial intelligence. Our research group has a strong track record of conducting cutting-edge research and has published numerous papers in top-tier conferences and journals. We encourage interested students to visit our Projects and Publications pages to learn more about our previous research.

As a member of our research group, you will have access to a self-motivated and flexible working environment that promotes efficient research. We provide the necessary training and guidance to help our students become independent researchers and experts in their chosen fields. Additionally, our research group actively collaborates with foreign researchers, providing our students with exposure to diverse perspectives and networking opportunities. Upon graduation, our alumni have a strong network and receive numerous offers in academia, research labs, and the industry. You can find more information on Join Us page.

If you are interested, please send your CV/resume and transcript to ina.interview@gmail.com and dhan.ee@kaist.ac.kr.

For students in Graduate School of AI, please contact to yechankim0502@gmail.com (Yechan Kim, GSAI Master Student).

Undergraduates (Individual Study, URP)

INA research group is currently seeking undergraduate research interns (상시 채용). If you are interested in conducting research on any of the topics below, please send us your resume and transcript to ina.interview@gmail.com.

1. Distributed Training at Scale (ML Systems)

Recently, the scale of AI models has been increasing to improve accuracy in tasks such as image and text processing, and the amount of input data has also been increasing rapidly. To train these large-scale models and datasets, distributed learning using multi-GPU and multi-node is essential, and research to improve the efficiency of distributed learning is also actively being conducted. In our laboratory, we have researched a framework called TSPipe, which enables efficient Knowledge Distillation model training using multiple GPUs through model parallelization. Based on this research experience, we plan to conduct research in this URP project to accelerate distributed learning by analyzing and resolving fundamental bottleneck points in distributed learning using multi-node.

Contact: Hwijoon Lim (hwijoon.lim@kaist.ac.kr)

2. 3D Volumetric Video Streaming (Video Systems)

Through this URP project, we are researching the live streaming process of 3D videos from real-time 3D capture to encoding and distribution to viewers over the internet. With the advent of the metaverse era, immersive content such as VR and AR and 3D volumetric videos are gaining attention. While video on demand streaming of 3D videos has been researched so far, live streaming is an unexplored area. Based on our LiveNAS research experience, which utilized deep learning to improve traditional 2D live video streaming, our laboratory aims to achieve efficient live streaming of 3D videos. We plan to convert the inefficient representation and transmission method of existing 3D data that does not meet the live streaming conditions into a lighter form and conduct research on real-time encoding and transmission.

Contact: Jaehong Kim (jaehong950305@gmail.com)

3. Cost-effective Retrospective Video Analytics (Video + ML System)

Retrospective video analytics refers to an application that sends a query such as “how many buses are in this video?” to a database-stored video and receives a response. With the recent development of deep learning, there has been an explosive increase in demand for retrospective video analytics (for example, in traffic/city planning, business/sports analytics, and healthcare), and many studies have been conducted to optimize it in terms of computing cost. In this research, we will study how to improve the efficiency of the system by optimizing the video encoding method and joint optimization between the ML pipeline and video encoding for retrospective video analytics. Through this research, we can learn about knowledge and recent trends related to retrospective video analytics, image classification, and video encoding, and have the opportunity to experience system research directly by developing large-scale systems.

Contact: Juncheol Ye (juncheolye@gmail.com)