Myeongjun Kim Personal Statement

After a project I could not finish, I read the leftover cloud bill line by line.

On December 5, 2024, AWS charged me ₩134,772. A month later another ₩94,138 went out, close to ₩230,000 in total. In a university team project I was responsible for the back end and infrastructure, but with my skills at the time I never managed to connect the front end to the server. After the project ended, a load balancer, servers, disks and public IPs were still running in the Sydney region. I was shaken, but I started by reading the billing details, found the leftover resources and deleted them. AWS Support asked for a monthly breakdown, the cause and measures to prevent it from happening again; I wrote and submitted these myself, asking GPT along the way. I did not stop at removing the problem: I also wrote down what to check so the same charges would not come back. This was when I first learned that building code and taking responsibility for every resource left when a service is shut down are different jobs. I have not confirmed whether a refund was made, so I do not claim one.

I held together a team whose first meeting only two people attended, using issues and reviews, and saw the service through to the end.

On June 26, 2026, I created the CLOV repository for the four of us. I was the team lead and a full-stack developer. Only one teammate and I came to the first meeting, and the other three were all non-majors on their first team project. Before explaining anything, I built a prototype so they could see the features and screens. Each teammate used a different AI tool and the code often clashed, so I set up a flow: open an issue before starting work, check the code and how it runs before merging, and leave the reason for each change in a review. When spoken explanations did not stick, I wrote the basic concepts and the overall service structure in Markdown documents and applied the same standard again in the next issue. At first I became a bottleneck by reviewing every merge myself, but after repeated documents and reviews, teammates understood their own tasks and the direction of the service, and problems decreased. By August 4 we had 152 issues and 266 pull requests, 72 of which I reviewed. I set up Ubuntu, Nginx, Docker and HTTPS myself and finally connected the server at a real address, something I could not finish at university.

I stepped out of familiar development work and started over as a beginner in networking and security.

On August 11, 2026, I became a beginner again in the SKT ALEPH network security program. VLANs, routing, ACLs and pfSense were all new to me. Whenever a connection dropped in the VMware and pfSense labs, I narrowed down the cause by checking ping, logs, IP ranges and configuration values in that order. The network flow I had passed over without understanding in the CLOV deployment, this time I recorded not just the result but the order of my checks. On August 31 I started conversations with new teammates first, and on September 4 I helped a teammate next to me who was struggling to keep up with class. On September 13 I submitted a paper that tested, on 2,830,743 CICIDS2017 records, whether an intrusion detection model keeps its performance on attacks it has never seen. I am carrying the choice to step outside familiar development forward through attendance, daily rituals and reproducible research.