M.S. student at OCTOLAB (Distributed Space Systems Lab), Korea Aerospace University, working on machine learning for satellite constellations.

• Federated Learning over LEO Satellite Constellations — asynchronous FL under intermittent ground-satellite connectivity
• Onboard AI — deploying and updating ML models on resource-constrained satellite platforms
• Satellite Digital Twins — high-fidelity constellation simulation for learning-system design and validation

FedSAP (Satellite-Adapted Federated Learning) — a federated learning framework that treats LEO satellites as clients, with a theoretically derived optimal global learning rate for staleness-aware aggregation. Validated on a 238-satellite Walker-Delta constellation against FedAsync, FedBuff, FedSpace, and FedOrbit. Manuscript in preparation — code will be released upon publication.

• M.S., AI & Space Systems Convergence — Korea Aerospace University (advisor: Prof. Zizung Yoon, OCTOLAB)
• B.S., Software Engineering — Korea Aerospace University

Python · C · C++ · Kotlin · JAVA · Javascript · MATLAB

PyTorch · Pysyft · Federated Learning (Flower / custom simulators) · LaTeX · MATLAB · Docker · JIRA · Confluence · Slack

• Email: taekhyun.kim@kau.kr
• CV: .