JoCoding AX Partners — Teaching Assistant
Teaching assistant for Claude Code-based AX training, answering questions during hands-on sessions and helping with environment setup.
I enjoy building and running software, and tracking down the cause and fixing it when something goes wrong. I work across web services, cloud infrastructure, systems programming, and open source.
At a university lab, I developed and operated DCU Code, an online judge service with 4,000+ registered users, and served as SWLAB team lead as an undergraduate. At 42 Gyeongsan, I did C/C++ systems programming projects and took charge of the infrastructure as tech lead of a five-person real-time web game team; most recently, as team lead and presenter at the 42.Zip contest, I built a work platform in three days and won the Silver Award among 13 teams. In a project-based internship, I implemented the backend and automated VM provisioning of a cloud migration portal, and to fix problems I ran into on my own devices, I also contributed to Linux-related open source, including systemd.
In recent team projects, I fed the design we worked out as a team into AI coding tools to generate the implementation, then refined it to fit each feature's purpose and the users' needs. Rather than tying myself to particular technologies, I care about learning what's needed and turning it into real results.
Teaching assistant for Claude Code-based AX training, answering questions during hands-on sessions and helping with environment setup.
Reviewing LLM-generated output and selecting whether it matches the expected result. No separate grading rubric — judged output by output; results feed into model training (tuning) data.
In an AI-driven cloud migration service enhancement project, owned the API-based modernization of the existing management portal. Implemented a migration wizard (automatic inventory collection, environment and network selection, progress screens) and the FastAPI backend, and automated OpenStack VM provisioning with Terraform, including background execution, real-time logs, and automatic cleanup on failure. Helped design the pub/sub structure of a C++ agent installed on each VM to support communication and control.
Developed and operated DCU Code, an online coding education service, for about two years (Vue.js, Django): built a Judge0-based code execution result feature, fixed bugs at Jira-ticket level, and led development of an OpenAI API-based AI coding assistant. Held, as an undergraduate, the team lead role usually filled by graduate students, balancing hands-on development with task allocation and coordination. Took part in migrating the Docker-based service to Kubernetes, tracing restart-looping and resource-hungry Pods through their status and logs to resolve them. Developed SWLAB, an isolated per-student practice environment.
Trained and experimented with a deep-learning-based computer vision model for manufacturing defect detection, using PyTorch, OpenCV, etc.
Main Projects
Entered Gyeongsan Innovation Academy's "42.Zip" contest (13 teams) and won the Silver Award. Interview requests and meeting room bookings were handled through Slack messages, so trainees couldn't easily see available times and staff had to check and adjust overlapping requests by hand; improved this with a web platform that combines request → staff review and adjustment → confirmation into a single flow. Separate screens for trainees and staff, a mobile PWA, and confirmation notifications. As team lead and presenter, led the three-day build and owned the backend: worked out the design with my teammate and fed it to AI coding tools to generate the implementation and refined the result to fit the feature's purpose and user needs.
A multiplayer web game where falling words are typed in real time, with lobbies, spectator mode, and in-game chat — the final team project of the 42 Gyeongsan Common Core. Served as tech lead of a 5-person team, coordinating development collaboration. Implemented the backend game/lobby/auth modules and the real-time game communication (WebSocket/Socket.IO), plus JWT + 42 OAuth authentication. Owned the infrastructure: a Docker Compose development environment, Kubernetes deployment manifests, an nginx WebSocket proxy, Prometheus + Grafana monitoring, and CI built on repository_dispatch that automatically updates submodule pointers. Monitoring grew out of learning about a DCU Code outage only after it happened: service state is collected as metrics, with alerts when thresholds are exceeded. Contributed to the structure of the React frontend's pages, types, and hooks.
Improved a legacy jQuery/WebSocket-based cloud migration portal toward an API-centric design, implementing user workflows and management features. Built management screens and backend features such as checking migration target inventory, selecting environments, and showing task progress. Since target VMs need to be created automatically for a truly convenient migration, implemented Terraform-based OpenStack VM provisioning automation along with asynchronous task execution, log viewing, and failure handling.
To remove the hassle of setting up a development environment for every lab exercise, developed as a team a Kubernetes-based sandbox system that gives each student an independent practice environment reachable from a web browser. In the per-user Pod allocation and SSH-based practice environment, built the KubeSSH authentication (backed by the DCU Code database) and the SSH access page, and verified and troubleshot the Pod environment.
Related paper: Per-student sandbox practice system
A 3-person capstone design team project forked from DCU Code. An online coding learning platform offering automatic grading of textbook exercises and public problems, a community board, user rankings, group study, and real-time video calls. Responsible for a ChatGPT API-based AI coding assistant (see DCU Code); GitHub auto-push integration for accepted source code; and service UI and branding changes (logo, font, color theme). Worked on both the Vue.js frontend and backend.
A project improving and operating SNSLAB's existing online judge service. 4,000+ registered users; students write and submit code on the web and check automatic grading results. Improved and maintained the Vue.js frontend and Django backend and fixed bugs. Built a Judge0-based code execution result feature so students can run code right away without opening a separate program or site. Led development of a ChatGPT API-based AI coding assistant that explains the cause of an error without showing the fixed code, so students can find it themselves, built to cut questions to teaching assistants and help students who hesitate to ask. Took part in operating the Docker-based deployment environment and its migration to Kubernetes, tracing restart-looping and resource-hungry Pods through status and logs.
Related papers: Anti-cheating features AI coding assistant
42 Gyeongsan Industry Partnership Projects (Valeo Mobility Korea)
42 Common Core Projects
Other Projects
Main Education & Degrees
Joined to deepen my grasp of programming fundamentals and how systems work after building web services in a university lab. A project-based, peer-learning software program of the 42 Network: no fixed lectures or professor-led classes — learning by building projects and reviewing and evaluating them with peers. Hands-on C/C++ memory management and systems programming, UNIX processes and IPC, network socket programming, algorithms, and Docker-based service setup. Completed Common Core through individual assignments and peer evaluation; as tech lead of the final team project ft_transcendence, built the real-time game backend and the deployment/monitoring environment.
Completed Training Programs
Overseas Programs
Google Scholar · DBpia author page
First author
Co-author
| Language | Level |
|---|---|
| Korean | Native |
| English | Intermediate (OPIc IM2) |