Minsu Kim


Minsu Kim

polynomial6 AT gmail.com
yet another yak shaver

Interests

MLOps Infrastructure

Interested in efficient and maintainable designs and solutions needed for serving large Deep Learning models in production.

Algorithm Design

I’m interested in designing & analysing efficient algorithms for specific problems. I’ve written thousands of C++/Java codes for numerous algorithms and data structures solving many well-known problems. Especially, I like to participate in competitive programming contests where the ability to write correct and efficient program in a limited time is important.

Work and Research Experiences

Google (Mountain View) 2022.9-Current

Software Engineer

  • Improving Google Search Ranking with Deep Learning Models.

Google Korea 2019.3-2022.9

Software Engineer

  • Working as a Search Feature developer, I took ownership for the features in both Mobile and Desktop with the millions of daily active users. The development process included triggering, serving, rendering across the Google’s technical stacks, and evaluation running some experiments and working on a user metric analysis based on them as well.

Vision and Learning Laboratory, Seoul National University 2016.7-2016.8

Undergraduate Research Intern

Worked on improving the accuracy of movieQA, in which it is necessary to train the deep neural network via various sources like sentences and video feature.

Education

Seoul National University 2010.3-2017.8

B.A. in International Relations and Computer Science and Engineering

  • Overall GPA : 3.77/4.3
  • GPA of CSE courses : 4.2/4.3

Korea Advanced Institute of Science and Technology 2017.9-2019.9

M.S. in School of Computing

Publications

MC-SDN: Supporting Mixed-Criticality Scheduling on Switched-Ethernet Using Software-Defined Networking Proceedings of RTSS ’18

Kilho Lee, Taejune Park, Minsu Kim, Honsung Chwa, Jinkyu Lee, Seungwon Shin, Insik Shin

  • Helped implementing a case study in an autonomous driving car, scaled down to 1/10th. Based on a F1/10 competition model, I implemented a sensor fusion that utilizes the sensor data of a depth camera and a LiDAR, with PID control.
  • During this, I also made a simple bug fix in the network of ROS, which made UDP transport in ROS much more reliable. (ros#957)

Awards and Honors

ACM-ICPC Daejeon Regional Main Contest 2015.12

ranked 7th place(team EeveE)

Codeforces

rating : 1930(7% of users, handle : minsu)

Skills

Programming Languages

  • C/C++, Python, Java, C#

Other Tools

  • LaTeX, Markdown, Vim, Sublime Text, Git