Skip to content

Presentation

Dissertation defence deck and supporting materials. Clicking a deck below launches the real reveal.js presentation in full — keyboard shortcuts, chalkboard, navigation menu, all of it. Use F for fullscreen, S for speaker view, O for slide overview, ? for the full shortcut list.

Live decks

  •   Defence v3 — Consolidated (current)


    Renovated 2026-04-18 · reveal.js · 24 content slides + 3 Q&A backup

    Three-act arc: The Blind Spot → The Evidence → The Answer. Opens with the punchline (5× asymmetry) and walks backward through nine evidence blocks organised by type rather than by paper. Consolidates all materials from this site into one narrative.

    Open defence v3   Source qmd

Supporting materials

The scientific summary of every paper referenced in the deck lives in the Papers tab. The peer-review trackers for papers currently in revision live in Reviews. The methodology for how these papers were written lives in Workflow. The deck pulls from all three.

Deck structure at a glance

Act 1 — The Blind Spot (5 slides)

  • 국문 초록 · The Result to Remember (the teaser, 5× asymmetry) · Why the Blind Spot Exists · The Research Question + Portfolio Arc

Act 2 — The Evidence (9 slides, Setup / Finding / Implication)

  • Centrifuge (J1+J3) · Numerical spine (J2, J11) · Op³ integration · Probabilistic capacity (J5) · Field record (V) · Why cointegration fails (V) · The feature that works (B) · Cross-soil transfer (E) · Three-channel Bayesian fusion (A)

Act 3 — The Answer (8 slides)

  • What Changes for Engineers · Limitations · Contributions · Future Work · One Thing to Remember (callback) · 감사합니다

Anticipated Q&A backup (3 slides) — Methodology probes · Scope & generalisation · Downstream impact

Which paper appears in which slide

Every paper in the portfolio is represented in the deck. This table answers "where do I watch the J5 story?" for a committee member or reviewer who wants to jump to a specific paper's slide in the rehearsal recording.

Paper Scientific summary Deck slide Act Role in deck
J1 Year-1 centrifuge (published 2025) 7 Act 2 · Evidence 1 Shares the centrifuge slide with J3 (T1–T3 dry-sand baseline)
J2 3D FE-calibrated Winkler 8 Act 2 · Evidence 2 Numerical spine — the −3.9 % field match + power-law headline
J3 Centrifuge saturation & backfill 7 Act 2 · Evidence 1 Shares centrifuge slide with J1 (T4–T5 saturated extension)
J5 Monte Carlo capacity 10 Act 2 · Evidence 4 Probabilistic capacity — 1,794 realisations, 30 % 5th-pct gap
J11 Vesic dissipation 8 (backbone) Act 2 · Evidence 2 Theoretical backbone for J2; cited as mechanism-basis footnote on the J2 slide
V 32-month field + state-function 11 + 12 Act 2 · Evidences 5 & 6 Field record on slide 11; cointegration-failure + equivalence theorem on slide 12
B Buckingham-Pi features 13 Act 2 · Evidence 7 Feature-selection slide — coherence wins, fixity ratio fails
Op³ Framework software paper 9 Act 2 · Evidence 3 Integration enabler — 39 benchmarks, PyPI, makes J5 computationally feasible
A Bayesian decision 15 Act 2 · Evidence 9 Capstone — three-channel Bayesian fusion + VoI
E Physics-informed encoder 14 Act 2 · Evidence 8 Cross-soil transfer — CPT-conditioned latent space

Bookkeeping checks the table enables

  • Is any paper orphaned? (Every paper should have at least one slide. If the table has a blank slide column, the paper is missing from the deck and either needs a slot or the deck's scope needs to be reconsidered.)
  • Is any slide overloaded? (A slide covering two papers — e.g. slide 7 for J1 + J3, slide 12 for V Part A + Part B — should read as a single claim, not two stitched half-claims.)
  • Does the Q&A backup cover all reviewer tracker items? Cross-reference against J2 R2, J3 R1, and V-MSSP trackers.

Why this structure

The PhD committee remembers the spine of the argument, not the details. Three audience build-ups are engineered explicitly:

  1. Punchline first (slide 3). The five-fold asymmetry appears as a teaser before the problem is introduced. Committee hears the destination in the first 90 seconds and stays oriented.
  2. Setup / Finding / Implication pattern on every evidence slide. One setup sentence, one quantified finding, one implication. No multi-claim slides.
  3. Callback closure. The final takeaway slide quotes the teaser verbatim — the committee feels the argument has come full circle rather than trailing off.

What lives under the hood

  • Source: presentation-source/defence_v3.qmd in this repo. Quarto revealjs with a custom theme (themes/op3_reveal.scss), chalkboard and menu plugins, title-slide background, and speaker notes per slide.
  • Rendered: docs/presentation/defence_v3/index.html (≈ 22 MB including 13 MB of figures, 8.8 MB of reveal.js runtime libraries, and the theme).
  • Auto-render: any push to presentation-source/** triggers .github/workflows/render-deck.yml, which re-renders the deck and commits the staged output back. For local runs: bash scripts/render_deck.sh.
  • Legacy deck: defence_v2 is archived at F:\TREE_OF_THOUGHT\PHD\presentation\ — not served on this site; v3 supersedes it.

Controls

Inside the deck: / to navigate, S speaker view, O overview, B chalkboard, M menu, F fullscreen, ? help, Esc exit overview. Alt+Click any element to zoom in.

Acknowledgements

KEPCO · MMB · Unison for the industrial partnership and site access. KAIST 70 g centrifuge team for the experimental facility. Advisor Prof. Sung-Ryul Kim for three years of guidance. SNU Civil and Environmental Engineering community.