A showcase of projects I've worked on and those I'm currently working on.

Led development of a content curation platform for media and playlists within a 6-person team. Built real-time chat with WebSocket/STOMP and resolved distributed session issues using Redis Pub/Sub; Boosted performance by migrating watcher count updates to Redis, significantly reducing DB load.
Improved search response times by 15% through refactoring QueryDSL to OpenSearch and JMeter testing.

Built an AWS news community platform with a team of 6. Implemented async user tracking (Spring Events, MongoTemplate) and a read-only MongoDB model. Automated nightly S3 backups with Spring Batch. Engineered a CI/CD pipeline (GitHub Actions, Docker, ECS), reducing deployment time by 95%. Provisioned supporting AWS infrastructure (IAM, RDS, S3).

Developed SonoCraftAR, an AR HoloLens system that enables DHH users to author personalized shape-based visualizations. Implemented real-time audio processing and script generation/compilation in Unity/C#, GPT API (4o); Designed a multi-agent AI pipeline, delegating specific tasks (e.g., prompt enhancement, code generation, code checker).
Co-authored and published a workshop paper at ISMAR IDEATExR 2025.

Led a 5-person team to engineer a financial analysis dashboard. Developed core backend APIs for tracking index performance, moving averages, and rankings. Optimized data processing by implementing a Java sliding window algorithm and complex PostgreSQL queries (aggregations, window functions), significantly reducing query latency.

Developed a dialogue-driven visual novel where you play as a sentient crown choosing a kingdom’s ruler based on persuasive arguments; leads to 3 different endings & routes. Implemented an event-driven system and the core game mechanics. Made in ~two weeks for the Pirate Game Jam 2025 with a team of 3.

Built the main user-facing pages for ImpactEducation with React and Tailwind CSS. Added English–Korean language support through a localization library and ensured consistent, responsive design across platforms.

Developed a CNN model for trash classification, fine-tuned on ~5000 images from HuggingFace with VGG16, deployed via a Flask web interface with HTML/CSS, achieving 75% accuracy and winning 3rd place in the EduHacks hackathon 2025 with a team of 4.