• Led graduate research on EMR-based nursing surveillance decision support and diagnostic classification.
• Built end-to-end modeling pipelines using KM-BERT ensembles, XGBoost, and both structured and text data.
• Explored conversational AI topics, including empathetic dialogue generation with AI-based feedback.

• Implemented evaluation-related code in a human-centered multimodal AI project.
• Bridged evaluation requirements with actual code and reviewable deliverables.

• Contributed to an NRF-funded clinical AI project centered on nursing surveillance decision support using EMR data.
• Implemented workflows for clinical text understanding, including keyword extraction, dependency parsing-based preprocessing, topic modeling, and similarity analysis.
• Provided interpretable results and web-based analysis interfaces for collaborative researchers.
• Achieved 92%+ diagnostic prediction accuracy in a multi-label task by ensembling text and structured data models (KM-BERT, XGBoost).

• Held multiple leadership roles in the official university programming club and served as president in 2022.
• Planned and led study groups on machine learning, big data, financial ML, and GNNs.
• Supported campus learning communities and resource sharing as part of the Notion Campus Leader program (24-2 ~ 25-1).

• Participated in an industry-led intensive program focused on cloud, CI/CD, security, and DevOps practices.
• Contributed to implementation tasks with a focus on security and DevOps roles in project-based training.

• Nursing surveillance required diagnosis-related classification, but key clinical signals were fragmented across heterogeneous EMR sources.
• Trained two KM-BERT models independently and averaged raw logits to stabilize text representation.
• The final model was reviewed for practical classification behavior rather than a single standalone score, including rare-class recall and available-data constraints.

Evaluation Scope Balanced · Available Data Core EMR · Rare-class Recall High

• ML experiments and deployment work often scatter metadata, records, and artifacts across tools, making reproducible local observability hard to maintain.
• Used a `.contexta/` workspace as the home for separated metadata, records, and artifact storage.
• Implemented a local observability foundation for consistently managing and inspecting ML execution history and artifacts.

Architecture Local-first · Contract Schema-first · Workflow Trace / Compare / Recover

• Existing Python graph libraries and generic dataframe wrappers often struggle to combine memory efficiency, traversal performance, and lazy query optimization for large graph analytics.
• Designed GraphFrame to own Arrow RecordBatches directly.
• Established the foundation for a graph analytics engine with Arrow columnar memory, CSR-based traversal, and lazy collect execution.

Neighbor lookup O(degree) · Memory model Arrow · Execution Lazy collect

• The product needed ML components that could recognize user emotion and support a diary-generation workflow from daily conversation logs.
• Centered the ML workflow around KoBERT-based emotion recognition and connected it to chatbot and summarization flows.
• The project delivered an MVP-level conversational diary experience with an emotion recognition pipeline.

Pipeline Korean NLP

• Developers working alone often need UI, performance, or code quality feedback, but they rarely have an easy way to gather role-specific input at the right time.
• Separated role constraints and contextual input so responses stayed scoped instead of generic.
• The prototype showed how structured prompting can make LLM feedback more useful for project review.

Feedback modes Role-based

• The product needed ML components that could classify emotional polarity and generate stylistic variations of text to support richer blog content workflows.
• Used KoBERT for polarity recognition and KoBART for style transfer.
• The project gave the team practical AI modules for blog-oriented text processing.

Model scope Classification + Generation

• The project needed cloud infrastructure and platform foundations that could reliably support search, notifications, admin workflows, and ongoing operations.
• Organized CI/CD and cloud foundations so the React and Spring service could fit into a deployable, monitorable workflow.
• Built a practical service foundation for a closed internal mileage commerce platform.

Foundation CI/CD + Cloud