Home
On Mon September 09, 2024

Speaker

한기용

Title

젊은 날의 나에게 들려주고 싶은 이야기

Abstract

대학원 졸업 후 한국에서 5년, 실리콘밸리에서 개발자/리더로 24년 일하면서 배운 다양한 커리어/삶의 교훈들을 학생들과 공유하고자 합니다

Bio

• UpZen 창업자 (2024-)

• San Jose State University Applied Data Science 대학원 강사 (2024-)

• 엔젤투자자/컨설턴트/어드바이저 (2016 - ):
  ◦ 엔젤투자: 굿타임, 차트메트릭, 셀렉트스타 등 10여개의 회사
  ◦ 한국 대기업 컨설팅: SK텔레콤, 현대카드, SK아카데미, 이마트 등등
  ◦ 스타트업 어드바이저: 몰로코, 블라인드, 쿼리파이, 월급쟁이부자들 등등

• 스타트업 액셀러레이터 멘토: 500 Global, Techstars, 롯데벤처스

• Udemy 시니어 디렉터 (2014 - 2018)

• Yahoo 엔지니어링 디렉터 (2004 - 2011)

• 삼성전자 엔지니어 (1995 - 2000)

• 서울대학교 컴퓨터공학과 학사/석사

Language

Korean

On Mon September 23, 2024

Speaker

Katerina Argyraki

Title

Internet Transparency and Neutrality

Abstract

The overarching theme of my talk will be Internet transparency. I will start from two fundamental questions: is it possible to infer a network’s neutrality based on external observations? is it possible to localise neutrality violations to specific network areas based on external observations? I will give the answers, then describe how we use them to reason about ISPs’ traditional traffic-differentiation practises (policing or shaping traffic from specific content providers). Then, I will argue that there is a more surreptitious (and more dangerous?) form of differentiation on the rise: in-network caching; and I will discuss whether caching — and the current Internet architecture, which relies heavily on it — is fundamentally incompatible with the concept of neutrality. I will close on a more fun note: how to improve Internet transparency by extracting thousands of Internet performance-metrics every day from public game-streaming footage.

Bio

Katerina is an associate professor of computer science at EPFL, where she does research on network architecture and systems, with a particular interest in network transparency and neutrality. She received an IRTF applied networking research prize (2020) and Best Paper awards at SOSP (2009) and NSDI (2014), all shared with her students and co-authors. She has been honored with the EuroSys Jochen Liedtke Young Researcher Award (2016) and three teaching awards at EPFL. Prior to EPFL, she worked at Arista Networks from day one, and received her PhD from Stanford (2007).

Language

English

On Mon September 30, 2024

Speaker

Marco Canini

Title

Programmable Networks for Distributed Deep Learning: Advances and Perspectives

Abstract

Training large deep learning models is challenging due to high communication overheads that distributed training entails. Embracing the recent technological development of programmable network devices, this talk describes our efforts to rein in distributed deep learning's communication bottlenecks and offers an agenda for future work in this area. We demonstrate that an in-network aggregation primitive can accelerate distributed DL workloads, and can be implemented using modern programmable network devices. We discuss various designs for streaming aggregation and in-network data processing that lower memory requirements and exploit sparsity to maximize effective bandwidth use. We also touch on gradient compression methods, which contribute to lower communication volume and adapt to dynamic network conditions. Lastly, we consider how to continue our research in light of the enormous costs of training large models at scale, which make it quite hard for researchers to tackle this problem area. We will describe our ongoing work to create a new approach to emulate DL workloads at a fraction of the necessary resources.

Bio

Marco does not know what the next big thing will be. He asked ChatGPT, though the answer was underwhelming. But he's sure that our future next-gen computing and networking infrastructure must be a viable platform for it. Marco's research spans a number of areas in computer systems, including distributed systems, large-scale/cloud computing and computer networking with emphasis on programmable networks. His current focus is on designing better systems support for AI/ML and providing practical implementations deployable in the real world. Marco is an Associate Professor of Computer Science at KAUST. Marco obtained his Ph.D. in computer science and engineering from the University of Genoa in 2009 after spending the last year as a visiting student at the University of Cambridge. He was a postdoctoral researcher at EPFL and a senior research scientist at Deutsche Telekom Innovation Labs & TU Berlin. Before joining KAUST, he was an assistant professor at UCLouvain. He also held positions at Intel, Microsoft and Google.

Language

English

On Mon October 07, 2024

Speaker

최재호

Title

생성형 AI와 검색의 진화

Abstract

최근 생성형 AI(Generative AI)의 발전은 학계와 산업 전반에서 혁신을 이끌며, 특히 정보 검색 (Information Retrieval) 분야의 기존 패러다임을 근본적으로 재편하고 있습니다. 생성형 AI는 자연어 이해, 문맥 인식, 사용자 의도 파악, 시맨틱 랭킹 등의 능력에서 기존 알고리즘을 능가하며, 사용자가 보다 적은 노력으로 유의미한 결과를 얻는데 중요한 역할을 하고 있습니다. 지난 20여 년간 독자적인 검색 엔진을 개발해 온 네이버는 생성형 AI를 적극적으로 도입하여, 사용자 의도와 맥락을 더욱 정확하게 파악하고, 개인 맞춤형 결과를 제공하는 방향으로 진화하고 있습니다. 이를 통해 단순한 키워드 매칭 기반의 정보 검색을 넘어 정보 탐색의 효율성을 극대화하고, 복잡한 질문에 대해 종합적인 답변을 제공하여 사용자의 만족도를 높이고 있습니다. 뿐만 아니라, 사용자들의 검색 경험을 개인화된 피드(Feed)로 확장하여 새로운 정보 접근성의 가능성을 열어 가고 있습니다. 이번 강연에서는 생성형 AI를 바탕으로 발전하고 있는 네이버 검색에 대해 소개하고 개발 과정에서 얻은 인사이트를 공유하고자 합니다.

Bio

최재호 부문장은 네이버에서 AI 검색과 추천을 담당하는 발견/탐색 프로덕트 부문을 총괄하고 있습니다. 서울대학교 건축학과를 졸업하였고, 미국 매사추세츠 대학교(UMASS)에서 정보 검색 분야로 석사 학위를 받았습니다. 2003년 네이버에 입사하여 통합검색을 비롯한 다양한 검색서비스와 AI 추천시스템 AiRS (AI Recommender System)를 개발하였습니다. 최 부문장은 정보 검색과 AI 추천 분야에서 깊이 있는 전문 지식을 보유하고 있으며, 네이버의 혁신적인 서비스 개발에 기여하고 있습니다.

Language

Korean

On Mon October 14, 2024

Speaker

이미경

Title

그린 리더가 세상을 바꾼다

Abstract

본 강연에서는 제가 환경운동가로 일해 온 활동에 대해 발표하고, 이를 통해 기후변화 문제가 인류의 문명 자체를 위협하는 일임을 인식하고 앞으로의 해결책에 우리 모두가 적극적으로 나서야함을 얘기하려고 합니다.

강의 목차
1. 환경재단 소개_도전의 역사
2. 기후변화 인식의 역사_ESG는 누가, 왜 어떻게 시작하게 되었나
3. 기술의 방아쇠와 표적_나는 왜 지금 이 공부를 하나
4. 창의적인 신기술의 탄생요건
5. 그린리더가 세상을 바꾼다

Bio

• 연세대 국문과/심리학과 석사
• 삼성사회봉사단, 프랭클린 코비 코리아
• 2002년 환경재단 사무국장, 사무총장, 상임이사, 대표
• 환경부 중앙정책위원회 위원, 탄소중립위원회와 수소경제위원회 위원
• 현재 하이브, 삼성SDI 사외이사

Language

Korean

On Mon October 28, 2024

Speaker

우신애

Title

실리콘 밸리에서의 삶: 스타트업부터 NVIDIA까지

Abstract

대학원 시절 인턴 및 Visiting Scholar부터, 졸업 후 실리콘밸리에서 스타트업 및 대기업을 거친 다양한 커리어 경험들을 학생들과 공유하고자 합니다.

Bio

- NVIDIA Senior Engineer (2024-)
- 스타트업: Alkira Inc (2021-), Nefeli Networks (2018-)
- 박사 과정 중: Intel Research Intern 및 UC Berkeley Visiting Scholar
- UC Berkeley 포닥
- KAIST 학사/석사/박사

Language

Korean

On Mon November 04, 2024

Speaker

나태주

Title

AI시대의 시쓰기

Abstract

인공지능은 우리 삶 깊숙이 와 있고 미래를 사는 누구도 피할 수 없는 삶의 조건입니다. 현재 AI 글쓰기는 산문 문장은 비교적 진전이 있지만 운문 문장은 그렇지 않습니다. 그것은 산문 문장과 운문 문장의 차이 때문입니다. 산문 문장은 지적인 영역으로 지식이거나 지성이 산문의 글감입니다. 운문 문장은 감정의 영역으로 정신적 특성이며 변화무쌍한 것입니다. 나는 강연을 통해 AI 시대 시 쓰기의 방향을 모색하고자 합니다.

Bio

시인 ․ 카이스트 석학교수 , 1971년 <서울신문> 신춘문예 시 당선으로 데뷔, 1964년 2007년까지 43년 교직생활, 1973년 첫 시집 『대숲 아래서』출간 이래, 『그래,네 생각만 할게』까지 창작시집 51권 출간, 문학 저서 총 200여권 출간, 흙의문학상, 충남문화상, 한국시인협회상, 소월시문학상, 정지용문학상, 윤동주문학대상 등 수상, 공주문화원장 ․ 한국시인협회장 역임

Language

Korean

On Mon November 11, 2024

Speaker

Nicole Chen

Title

How to Integrate Econometric and AI Models for Traffic Safety and Human Factors Research?

Abstract

Road traffic injuries are predicted to rank as the fifth leading cause of death in 2030. Every day, over 3500 people die on the roads, which amounts to nearly 1.3 million preventable deaths and 50 million non-fatal injuries. Individuals, their families, and nations as a whole suffer considerable economic losses as a result of traffic injuries. The traditional approach to understanding traffic safety relies on crash data, which is reactive and has fundamental ethical and practical problems. A more effective approach to safety management requires evaluating the safety of locations over short time periods (e.g. minutes) to dynamically change the traffic environment and optimize safety in real-time. There are considerable opportunities for new technologies, such as connected and autonomous vehicles and advanced sensing and edge computing. A large amount of data also helps us better understand road user behaviour through AI. Moreover, human factors will remain crucial in the future of mobility systems. We need more advanced research on the safety of mixed traffic in the transition period and smart interaction between AV and vulnerable road users. Integrating econometric and artificial intelligence models is imperative in order to prepare for the future of road safety.

Bio

Dr. Tiantian Chen is an assistant professor at the Cho Chun Shik Graduate School of Mobility at KAIST. She received her PhD in Civil and Environmental Engineering from the Hong Kong Polytechnic University in 2021. She specializes in traffic safety, human factors, human-centered design, driving simulation, travel behavior, transport policy, and applications of statistical methods. She has over 20 journal publications in prestigious transportation & safety journals such as Transportation Research Part A, C, E, F, Transport Policy, Analytic Methods in Accident Research, and Accident Analysis & Prevention. She is now an Editor of the Case Studies on Transport Policy, and guest editor for the Special Issues in the International Journal of Sustainable Transportation, Multimodal Transportation, and Research in Transport Business and Management. Due to her expertise, she is also a member of the IEEE Emerging Transportation Technology Testing (ET3) Technical Committee. Additionally, she holds the positions of International Committee Chair at the Korean Society of Transportation (KST) and Chair of the Departmental Internationalization Committee. Her guidance and mentorship led her PhD student to secure the esteemed Best Paper (First Prize) at the 27th International Conference of the Hong Kong Society for Transportation Studies (HKSTS). Moreover, she was awarded the International Category Award Winner at the 2023 KAIST Q-Day (School level). In 2024, she was awarded the EWON Chair Assistant Professorship.

Language

English

On Mon November 18, 2024

Speaker

Eunjung Kim

Title

Flow-augmentation Technique and Its Applications

Abstract

We present a technique called the flow-augmentation and review how the technique can be used for designing graph cut problems. The flow-augmentation algorithm is a randomized polynomial-time algorithm, given a directed graph G, two vertices s and t, and an integer k, which adds a number of arcs so that for every minimal st-cut Z in G of size at most k, Z becomes an minimum st-cut in the resulting graph with probability 2^{-poly(k)}. The directed flow-augmentation tool allows us to prove fixed-parameter tractability of a number of problems parameterized by the cardinality of the deletion set, whose parameterized complexity status was repeatedly posed as open problems: (1) Chain SAT, defined by Chitnis, Egri, and Marx [ESA'13, Algorithmica'17], (2) a number of weighted variants of classic directed cut problems, such as Weighted st-Cut or Weighted Directed Feedback Vertex Set. Furthermore, leveraging the directed flow-augmentation technique, we obtain a complete dichotomy theorem for Boolean CSPs parameterized by the number of unsatisfied constraints. The talk is based on a series of joint work with Stefan Kratsch, Tomás Masarík, Marcin Pilipczuk, Roohani Sharma, Magnus Wahlström.

Bio

Eunjung KIM obtained her PhD in 2010 from Royal Holloway, University of London. After post-doctoral fellowship at LIRMM-CNRS in France, she was a researcher at CNRS (National Center for Scientific Research, France) between 2011 and 2023. Since 2024, she is an associate professor at KAIST. Dr. Kim contributed some important results in the study of algorithms using structural graph theory and width parameters. Among those, she is one of the researchers who invented the notion of twin-width in 2020, which is now used widely and considered an important tool for algorithms design, logic, structural graph theory and so on. She is one of the authors who proposed the flow-augmentation technique, which was essential for resolving some long open problems in parameterized complexity. She is awarded with the CNRS bronze medal, which is given to an early-career researcher in each field of basic science in France as a recognition of research excellence and leadership.

Language

English

On Mon November 25, 2024

Speaker

이상원

Title

FlashDB에서 VectorDB로

Abstract

본 강연은 크게 두 가지 주제로 구성되어 있다. 우선, 플래시메모리저장장치 기반 데이터베이스 연구의 동기와 FAST FTL, IPL, X-FTL, Share 등의 쓰기 및 읽기 최적화 기법을 소개하고, 비휘발성메모리 (Non Volatile Memory, NVM)를 통한 데이터베이스 쓰기 연산 가속 기술을 소개한다. 다음으로, 생성형 AI에서 검색증강생성(Retrieval-Augmented Generation, RAG) 및 VectorDB의 중요성을 설명하고, 유망한 주요 연구 주제와 연구 결과를 다룬다.

Bio

이상원 교수 (서울대 데이터사이언스대학원)는, 서울대학교 컴퓨터공학과에서 ’91, ’94, ‘99년 학사, 석사, 박사 학위 취득 후, 한국오라클과 이화여대를 거쳐, 2002년부터 성균관대학교에서 플래시메모리저장장치 기반 데이터베이스 최적화에 관한 연구를 수행했고, 2023년부터 서울대학교 데이터사이언스대학원에서 생성형 AI 분야의 핵심 기술 요소인 벡터 데이터베이스 및 관련 기술 분야를 연구 중이다. 2024년 현재 정보과학회 우수학술대회목록 개편위원장과 데이터 소사이어티 회장을 역임하고 있다.

Language

Korean

On Mon December 02, 2024

Speaker

Aurojit Panda

Title

Runtime Protocol Refinement Checking for Distributed Protocol Implementations

Abstract

Despite significant progress in verifying protocols, services that implement distributed protocols , e.g., Chubby or Etcd, can exhibit safety bugs in production deployments. These bugs are often introduced by programmers when converting protocol descriptions into code. In this talk I will describe a new technique we have been developing to identify these bugs at runtime: Runtime Protocol Refinement Checking} (RPRC). RPRC systems observe a deployed service's runtime behavior and notify operators when this behavior evidences a protocol implementation bug, allowing operators to mitigate the bugs impact and developers to fix the bug. We have developed an algorithm for RPRC and implemented it in a system called Ellsberg that targets services that assume the asynchronous or partially synchronous model, and fail-stop failures. We designed Ellsberg so it makes no assumptions about how services are implemented, and requires no additional coordination or communication. We have used Ellsberg with three open source services: Etcd, Zookeeper and Redis Raft.

Bio

Aurojit Panda is an assistant professor in the Computer Science department at New York University working on systems and networking. He received his PhD in 2017 from UC Berkeley, where he was advised by Scott Shenker. He has received several awards, including a VMware Early Career Faculty Award, a Google Research Scholar Award, an NSF Career award, best paper awards at EuroSys, SIGCOMM and OSDI, and a EuroSys test of time award.

Language

English

On Mon December 09, 2024

Speaker

Gwangsun Kim

Title

Breaking the Memory Wall: Near-Data Processing for Hyperscale Applications

Abstract

The memory wall has long been recognized as a critical challenge in high-performance systems, and it has recently become even more significant due to the exponential growth of machine learning model sizes. Meanwhile, recent advancements in interconnect technology, such as Compute Express Link (CXL), enable scalable memory system designs to address the memory capacity wall. Moreover, by offloading data and computation to CXL memory expanders to realize Near-Data Processing (NDP), the memory bandwidth wall can also be effectively mitigated. However, designing such a system should be done carefully, considering various design aspects that can affect the practicality of the solution.
In this talk, I will discuss key considerations and directions for building a practical NDP system architecture, including general-purpose computing, low-latency host communication, standard compliance, and cost-effectiveness. I will then present our recent work on an NDP architecture called Memory-Mapped NDP (M²NDP). M²NDP consists of two components: 1) Memory-Mapped Function (M²func), which enables low-latency host-device communication by addressing the overhead of conventional ring buffer-based task offloading, and 2) Memory-Mapped μthreading (M²μthread), a general-purpose, cost-effective NDP unit architecture that aims to maximize resource utilization by hybridizing CPU and GPU architectures. Finally, I will briefly outline future research directions based on the M²NDP architecture.

Bio

Gwangsun Kim is an Assistant Professor in the Department of Computer Science and Engineering at POSTECH. Previously, he was a Senior Research Engineer and Senior Performance Engineer at Arm Inc. He received the B.S. degrees in Electronic and Electrical Engineering and Computer Science and Engineering from POSTECH in 2010, and the M.S. and Ph.D. degrees in Computer Science from KAIST in 2012 and 2016, respectively. He has worked on various areas of computer architecture and systems, including memory systems, parallel architectures, GPU computing, systems for machine learning, near-data processing, networking, deep learning compiler, and simulation methodology. He is particularly interested in designing practical architectures for high-performance and scalable systems.

Language

English