- 𧩠Knowledge Engineer at Remember, designing ontology-based knowledge graphs and GraphRAG systems for HR data
- π¬ Research interests: representation learning, ontology induction from embeddings, mechanistic interpretability
- π M.E. in Computer Science and Engineering, Hanyang University (2026)
Peeling the Matryoshka: Extracting Ontological Structure from Retrieval Embeddings
EMNLP 2026 Main Conference Β· EunChan Kim, Inwhee Joe
Matryoshka embeddings encode a coarse-to-fine granularity gradient along their leading dimensions, validated across three model families. A 98K-parameter order-embedding probe renders that gradient as explicit is-a taxonomy DAGs over frozen retrieval embeddings, with no encoder retraining.
- π§ Remember, Knowledge Engineer (2026.07~)
- π© TIMWORK Inc., Machine Learning Engineer (2023.10~2026.06)
- πͺ The Matrix, Researcher (2022.03~2023.09)
- π₯ Deep In Sight, Research Engineer (2021.11~2022.03)
- π₯ Deep In Sight, Research Intern (2021.07~2021.10)
Python PyTorch Neo4j Cypher RDF/OWL LangGraph AWS Bedrock Docker
- π Books
- ποΈββοΈ CrossFit