Hello, I'm devcomfort.
I've been programming since 2013, building expertise across frontend, backend, databases, infrastructure, and the broader development landscape.
I'm currently pursuing a Master's degree in AI Convergence at GIST (Gwangju Institute of Science and Technology), where I'm actively engaged in AI research.

Current Research Focus: RAG (Retrieval-Augmented Generation), Multi-Agent Systems

I have experience with Tabular Learning through my undergraduate graduation research and Procedural Content Generation through the GIST CILab internship. Feel free to reach out regardless of the field—I'm always open to discussions.

Over the past 12 years of programming, I've continuously explored ways to innovate both UX and DX.
As a result, I've recently developed and maintained open-source libraries and CLI tools including @devcomfort/text-transcoder, blob-to-url, and Ubuntu Disk Toolkit.

• 🔭 Currently working as Tech Lead at (주)마시는친구들 (August 2024 ~ Present)
• 🎓 Master's Student in AI Convergence at GIST (September 2025 ~ Present)
• 🔬 Completed GIST CILab Internship (January-February 2025) - Co-authored IPCGRL paper as 2nd author
• 🌱 Currently learning: PyTorch, Graph Neural Networks, Reinforcement Learning
• 👨‍💻 All projects available at https://devcomfort.me/projects
• 📝 Blog posts: https://blog.devcomfort.me
• 💬 Welcome questions about career, development issues, and AI research
• 📫 Contact: im@devcomfort.me, devcomfort@t-eut.com

• GIST CILab Internship (Jan-Feb 2025): Participated in reinforcement learning-based game map generation research in the PCG team
• IPCGRL Paper Co-author: "Language-Instructed Reinforcement Learning for Procedural Level Generation" (IEEE CoG 2025 Accepted)
• Graduation Research: "Categorical Data Processing in the AI Era — Analysis of Effectiveness of Entity Embedding and Cat2Vec"

• Discovered training data bias
• Optimized data normalization pipeline: Reduced RAM usage from 502GB+ → 150GB, processing time from 2 hours → 1 hour

• combinatorial-config - Python library for automatically generating all combinations of experiment configurations. Useful for systematic hyperparameter exploration with Hydra and other configuration management tools.

• file_type_detector - Python library for detecting file types based on content analysis and file signatures.

• devcomfort-text-transcoder - Lightweight and efficient library for text encoding conversion in Node.js and browser environments. Supports multiple encodings including UTF-8, EUC-KR, and EUC-JP.

• is-html-element - Utility library for HTML element type checking and validation.

• ubuntu-disk-toolkit - CLI toolkit for Ubuntu disk management and operations (under development).

• chonkie-chunk-utils - Utility library for chunk processing and management (private repository).

TypeScript JavaScript Python

Usage: TypeScript/JavaScript for web frontend and backend development; Python for AI research, data analysis, and processing

Object-Oriented Functional Reactive

Approach: Extensible design using classes and interfaces; preference for functional programming (lodash, remeda); reactive programming for event streams and state management

React Svelte/SvelteKit SolidJS TailwindCSS

Focus: Web application development with emphasis on reactivity and performance

Node.js Express FastAPI Starlette

Usage: API development and server-side applications

PostgreSQL SQLite Prisma

Approach: Type-safe database access and modeling primarily using Prisma; PostgreSQL schema design and migration management; SQLite for local development environments

PyTorch JAX Flax Polars Pandas NumPy Matplotlib Seaborn

Usage: Model development and training (PyTorch); scalable ML systems (JAX, Flax); efficient data preprocessing and analysis (Polars, Pandas, NumPy); data visualization and experiment tracking (Matplotlib, Seaborn)

Docker GitHub Actions Vitest Pytest

Status: Docker usage; CI/CD pipeline setup using GitHub Actions; test writing with Vitest and Pytest

Tauri

Status: Interested in and learning Tauri for cross-platform development

• RAG (Retrieval-Augmented Generation) - Research on enhancing LLM capabilities through effective information retrieval and context augmentation
• Multi-Agent System - Research on designing and implementing collaborative AI agent systems for complex problem-solving

• Representation Learning - Research on exploring and validating efficient representation and transformation methods for various forms of data
  • Tabular Learning - Research on effective learning methodologies for tabular data
• Graph Neural Networks (GNN) - Effective modeling and learning of graph-structured data
• Explainable AI (xAI) - Research on ensuring explainability of AI models

• GitHub Showcase Competition Excellence Award (November 2024)
• TOPCIT (Test of Practical Competency in IT) - 557 points (Level 3 - Proficient) (May 2024)