INA Research Group

Welcome to Intelligent Network Architecture Lab @ KAIST!

 

INA research group pursues innovative ideas in/for Internet services and applications, cloud infrastructure, and systems that support artificial intelligence. We identify and anticipate new problems that arise from the evolution of Internet-/Cloud-based services and the development of new hardware, provide novel solutions for challenging problems in the real-world, design and implement the solutions in a way that reaches out for real-world impact. INA research group is part of the Computer Division of the School of Electrical Engineering.

INA research group provides a self-motivated atmosphere that allows for efficient research through flexible working hours. In the doctoral program, our research group provides the necessary training to become an independent researcher and become an expert in a particular field. There is also active exchange with foreign researchers, and Lab alumni have a strong network and receive many offers upon graduation, entering the industry, research lab, or academia as illustrated in Alumni@INA.

Open Position: We are looking for graduate students and undergraduate research interns to join the INA lab (상시 채용). We have several interesting topics such as large-scale deep learning, neural volumetric streaming, and retrospective video analytics. You can find more information on Join Us page.

   

Research

INA has published 14 papers at top-tier systems conferences, ranking 1st in Korea (refer to link). Currently, INA Lab is conducting research in three different directions, which are listed below. You can find a full list of projects and publications on Project@INA and Publication@INA.

Systems for AI and Cloud

  • Accelerating training for large-scale deep neural networks [ICML’22]
  • Accelerating DNA sequencing by learned index [Bioinformatics’22]
  • Optimizing microservice auto-scaling by graph neural networks [CoNEXT’21]
  • Optimizing systems for LLM training and inference

AI-augmented Media Delivery

Networked Systems (Data center, 5G) / Systems Security

     

News

May, 2024 Our papers titled “Accelerating Model Training in Multi-cluster Environments with Consumer-grade GPUs” and “TopFull: An Adaptive Top-Down Overload Control for SLO-Oriented Microservices” were accepted to SIGCOMM 2024. Congratulations Hwijoon, Jinwoo, Juncheol, Jaehyeong, Yongmok, and Hyunho!
May, 2024 Our paper titled “Scaling Beyond the GPU Memory Limit for Large Mixture-of-Experts Model Training” was accepted to ICML 2024. Congratulations Yechan, Hwijoon!
Apr, 2024 Our paper titled “Graph Neural Network-based SLO-aware Proactive Resource Autoscaling Framework for Microservices” was accepted to IEEE/ACM Transactions on Networking (ToN). Congratulations Jinwoo, Byungkwon!
Dec, 2023 Youngmok Jung successfully defended his Ph.D. thesis. Congratulations Dr. Youngmok Jung!
Jul, 2023 Our papers titled “Neural Cloud Storage” and “SAND” were accepted to HotStorage 2023. Congratulations Uitaek, Jinyeong!
Apr, 2023 Hyunho Yeo successfully defended his Ph.D. thesis. Congratulations Dr. Hyunho Yeo!
Feb, 2023 Our paper titled “AccelIR: Task-aware Image Compression for Accelerating Neural Restoration” was accepted to CVPR 2023. Congratulations Juncheol, Hyunho, and Jinwoo!
Jan, 2023 Jaehong, Yunheon, Hwijoon, and Youngmok won the Silver Prize (2nd place in Communication & Networks) in the 29th Samsung Humantech Paper Award.
Jan, 2023 Our paper titled “FlexPass: A Case for Flexible Credit-based Transport for Datacenter Networks” was accepted to EuroSys 2023. Congratulations Hwijoon, Jaehong and Inho!
Dec, 2022 We have two undergraduate research interns joining our group this winter. Welcome aboard Taehyun and Hyojin!
[More]
May, 2024 Our papers titled “Accelerating Model Training in Multi-cluster Environments with Consumer-grade GPUs” and “TopFull: An Adaptive Top-Down Overload Control for SLO-Oriented Microservices” were accepted to SIGCOMM 2024. Congratulations Hwijoon, Jinwoo, Juncheol, Jaehyeong, Yongmok, and Hyunho!
May, 2024 Our paper titled “Scaling Beyond the GPU Memory Limit for Large Mixture-of-Experts Model Training” was accepted to ICML 2024. Congratulations Yechan, Hwijoon!
Apr, 2024 Our paper titled “Graph Neural Network-based SLO-aware Proactive Resource Autoscaling Framework for Microservices” was accepted to IEEE/ACM Transactions on Networking (ToN). Congratulations Jinwoo, Byungkwon!
Dec, 2023 Youngmok Jung successfully defended his Ph.D. thesis. Congratulations Dr. Youngmok Jung!
Jul, 2023 Our papers titled “Neural Cloud Storage” and “SAND” were accepted to HotStorage 2023. Congratulations Uitaek, Jinyeong!
Apr, 2023 Hyunho Yeo successfully defended his Ph.D. thesis. Congratulations Dr. Hyunho Yeo!
Feb, 2023 Our paper titled “AccelIR: Task-aware Image Compression for Accelerating Neural Restoration” was accepted to CVPR 2023. Congratulations Juncheol, Hyunho, and Jinwoo!
Jan, 2023 Jaehong, Yunheon, Hwijoon, and Youngmok won the Silver Prize (2nd place in Communication & Networks) in the 29th Samsung Humantech Paper Award.
Jan, 2023 Our paper titled “FlexPass: A Case for Flexible Credit-based Transport for Datacenter Networks” was accepted to EuroSys 2023. Congratulations Hwijoon, Jaehong and Inho!
Dec, 2022 We have two undergraduate research interns joining our group this winter. Welcome aboard Taehyun and Hyojin!
Dec, 2022 Jeongmin Kim successfully defended his Ph.D. thesis. Congratulations Dr. Jeongmin Kim!
Nov, 2022 Our paper titled “Scalable and Secure Virtualization of HSM with ScaleTrust” was accepted to IEEE/ACM Transactions on Networking (ToN). Congratulations Juhyeng and Seongmin!
Sep, 2022 Our paper titled “Co-optimizing for Flow Completion Time in Radio Access Network” was accepted to CoNEXT 2022. Congratulations Jaehong, Yunheon, Hwijoon, and Youngmok!
Jun, 2022 Juhyeng Han successfully defended his Ph.D. thesis and will start a new career as a software engineer at SAP Labs Korea (which is the largest tech company in Germany). Congratulations Dr. Juhyeng Han!
May, 2022 Our paper titled “TSPipe: Learn from Teacher Faster with Pipelines” was accepted to ICML 2022. Congratulations Hwijoon and Yechan!
May, 2022 Our paper titled “NeuroScaler: Neural Video Enhancement at Scale” was accepted to SIGCOMM’22. Congratulations Hyunho, Hwijoon, Jaehong, Youngmok, and Juncheol!
Mar, 2022 Our paper titled “BWA-MEME: BWA-MEM emulated with a machine learning approach” was accepted to Bioinformatics’22. Congratulations Youngmok!
Feb, 2022 We have two undergraduate research interns joining our group this spring. Welcome aboard Jinyeong, Euijun!
Feb, 2022 We have two graduate students joining our group this spring. Welcome aboard Uitaek, Jaehyeong!
Feb, 2022 Hwijoon, Jaehong, and Inho won the Gold Prize (1st place in Communication & Networks) in the 28th Samsung Humantech Paper Award.
Jan, 2022 Byungkwon won Ph.D. Thesis Award, School of Electrical Engineering, for his dissertation “Enhancing Performance of Networked Systems in the Cloud by Considering the Characteristics of Resources and Workloads”.
Sep, 2021 Our paper titled “GRAF: a graph neural network based proactive resource allocation framework for SLO-oriented microservices” was accepted to CoNext’21. Congratulations Jinwoo!
Aug, 2021 We have one graduate student joining our group this fall. Welcome aboard Yechan!
Aug, 2021 We have three undergraduate research interns joining our group this fall. Welcome aboard Uitaek, Euijun, Jaehyeong!
Mar, 2021 Our paper titled “Towards timeout-less transport in commodity datacenter networks” was accepted to Eurosys’21. Congratulations Hwijoon, Youngmok!
[Less]