Academic website for Kyeong Sun Kim (김경선), PhD candidate in Civil & Environmental Engineering at Seoul National University (advisor: Prof. Sung-Ryul Kim; defense September 2026).
Live site: ksk5429.github.io/research
TMEF — Trust-weighted Multi-modal Evidence Fusion for Engineering Decisions.
How does heterogeneous evidence become trustworthy decisions?
Safety-critical engineering decisions draw on evidence that arrives in wildly different forms: continuous sensor streams, discrete lab results, expensive engineering simulations, expert judgment, AI-agent recommendations. Each source has its own reliability, cost, and failure mode. TMEF formalises this across four research questions:
- Unified representation of heterogeneous evidence — embed physically different signal types into a shared latent space without collapsing modality-specific structure.
- Trust-weighted fusion with source-aware credibility — Bayesian posterior with learnable, dynamic trust hyperparameters; handle disagreement, failure modes, consequence-dependence.
- Value-of-Information across cost-heterogeneous modalities — design the optimal next measurement when modalities differ in cost, time-to-result, and information gain per unit cost.
- Trust calibration for AI-mediated composite metrics — attribution, credit, and accountability for multi-modal AI decisions; when does the human know which parts of the input to trust?
My PhD dissertation is the structural health monitoring instance of this framework, validated on 22 centrifuge tests at the KAIST 70g facility, 1,794 Monte Carlo simulations, and 32 months of field monitoring on a 4.2 MW offshore wind turbine in Gunsan, South Korea.
Full program details: see the research page on the live site.
- Op³ — open-source Python framework for scour assessment of offshore wind turbine foundations.
pip install op3-framework· 140 tests · 39 V&V benchmarks · Zenodo DOI. - K-Fish — 9-agent LLM swarm with Bayesian confidence-weighted fusion and a Calibrator agent that audits its own reasoning. Brier 0.213 on 30 resolved Polymarket contracts.
- PhD dissertation — Integrated Numerical and Digital Twin Framework for Scour Assessment of Offshore Wind Turbine with Tripod Suction Bucket Foundations.
This repository hosts the source for ksk5429.github.io/research, built with the al-folio Jekyll theme (MIT).
- Content lives in
_pages/,_projects/,_news/,_bibliography/papers.bib, and_data/cv.yml. - Theme and layout files (
_layouts/,_sass/,_includes/,assets/) are al-folio upstream, modified minimally. - GitHub Actions workflow at
.github/workflows/deploy.ymlbuilds and deploys to GitHub Pages on every push tomain.
bundle install
bundle exec jekyll serveRequires Ruby ≥ 3.0 and Bundler. Then open http://localhost:4000/research/.
- Email: kyeongsunkim@snu.ac.kr
- GitHub: @ksk5429
- ResearchGate: Kyeong-Sun-Kim-2
Site source: MIT License (inherited from al-folio). Content (text, figures, publications): © 2026 Kyeong Sun Kim — all rights reserved unless otherwise noted.