JISUNG KIM
Open to opportunities · Worldwide

Curiosity.
Made tangible.

I’m Jisung Kim. I build research tools, explore data, and turn digital ideas into things you can try.

BIOMEDICAL ENGINEERING / YONSEI MIRAE
PORTFOLIO · 2026
01 / DIGITAL ARCHITECTUREHanyang ↘02 / RESEARCH SOFTWARERATRO ↘03 / DATA & DECISIONSSECOM ↘
SELECTED WORK / 01—03

Don’t just read.
Take a closer look.

Three different questions. Three working explorations. Open a model, move through a reference gait, or inspect the trade-offs in an analysis.

01 / DIGITAL ARCHITECTURE

Hanyang한양

An architectural world, down to the details.

HANYANG / MAIN HOUSE3D EXPLORER

Loads on demand · Main house 7.0 MB

A hanok-inspired asset collection, built to be explored and used.

Hanyang brings the house and entrance together. The DALBIT courtyard series expands the collection with modular gates, walls and props. Select an asset to inspect its geometry from every angle.

BlenderProcedural modellingGLB / FBXGodot / Unity
Project details & evidence

The viewer uses web-optimized versions of actual project GLB exports. The house and entrance are separate Hanyang LOD2 exports; DALBIT examples are LOD1 exports from the released courtyard pack. The poster depicts the assembled Hanyang compound, not a single exported component.

Models, textures, code and documentation were made with AI-assisted procedural workflows. The layout includes approximations from architectural references; this is not a measured historical reconstruction. Reference-rights review remains in progress.

The DALBIT courtyard pack’s itch.io release is verified. Unity and Fab submission preparation is not a store release. No sales, revenue or customer-adoption claims are made.

Drag to orbit. Pinch to zoom. Keyboard and buttons work too. Model files load only when you choose to explore.

02 / RESEARCH SOFTWARE

RATROREFERENCE MOTION

A question about movement. A tool for exploring it.

RATRO / PROTOTYPESCENE UNITS
MOTION INSPECTOR

Look beneath the movement.

Scrub one reference cycle, isolate the rig, and rotate your view. This is a small, working excerpt of the RATRO prototype.

HL STANCE
HR STANCE
FL STANCE
FR STANCE

Parametric reference animation. Arbitrary scene units. Not measured gait or validated anatomy.

Keeping the source, the signal and the unknowns in view.

RATRO connects individual and session records, layered 3D views, timelines, markers and analysis exports in a local research workspace.

ReactTypeScriptThree.jsPython / FastAPI
Prototype scope & limitations

This viewer reuses the prototype’s reference gait, rig definitions and transform calculations. The portfolio adds a simplified visual envelope and interaction controls; it does not expose private records or the application backend.

Scientific admission of external source models remains on hold. No external animal mesh is used in this excerpt. The visible skeleton and body envelope are illustrative, not an anatomical atlas or physics validation.

Evidence: implemented repository workflows and the 13 October 2026 source-model verification report. Implementation is AI-assisted. The repository is private.

03 / DATA & DECISIONS

SECOMQUALITY STUDIES

A better-looking metric is not the whole story.

INTERACTIVE / REVIEW CAPACITY

Review more. Find more?

Change the review budget and inspect three saved Decision Lab policies. Each circle represents one failure in the chronological test set.

314 test samples · 17 failures · chronological holdout

FAILURE CAPTURE MAPn = 17
CapturedMissed
62reviewed
3failures captured
14failures missed

59 passing samples also reviewed. Random review would capture an expected 3.36 failures.

Making the trade-off visible. Including the uncomfortable part.

Two separate studies of public semiconductor data: leakage-aware classification and chronological inspection-capacity analysis. These controls compare recorded results; they do not generate new predictions.

Pythonpandasscikit-learnEvaluation
Results, evidence & limitations

Frozen SECOM v1: 17 of 21 failures captured on a random holdout, at the cost of 145 false positives. Temporal sensitivity was weaker. This is separate from the chronological Decision Lab results above.

Decision Lab: the pipeline has run; interpretation remains Draft / awaiting review. One historical split and retrospective ranking do not establish operational advantage. The public dataset dates to 2008. No production-line or yield improvement is claimed.

Interpretation: Draft. No operational advantage established.

BACKGROUND / EXPERIENCE

Learning by
being there.

JUN — AUG 20268 weeks · completed

Korea Expressway Corporation

Yongin Branch · Customer Support Department

Internship-type public administration / office administration programme under the Ministry of Employment and Labor’s Youth Work Experience initiative. 29 June–21 August 2026.

Completion certificate reviewed
DEC 2025Started

Undergraduate research

Undergraduate research participation, beginning in December 2025.

Start date confirmed
YONSEIMIRAE CAMPUS

Biomedical engineering

Student at Yonsei University MIRAE Campus.

Existing profile
ABOUT / APPROACH

Open on the path.
Serious about the work.

I’m open to opportunities across countries, industries and company sizes.

Biomedical and life-science technology, semiconductors, audio and broader technology interest me. I’m exploring research, product, strategy, operations and software roles, through work I can demonstrate.

Data & evaluation

Python · pandas · scikit-learn

Interfaces & interaction

TypeScript · React · Three.js

Digital production

Blender · Godot · Git

Working practice

AI-assisted building · Testing · Documentation

Tools reflect reviewed project work. Interests are not claims of expertise across every field.

NEXT / LET’S TALK

What could we
make possible?

A role, a question, a useful problem. I’d like to hear about it. Availability and working arrangements can be discussed.

Connect on LinkedIn ↗