Skip to content
View mself8's full-sized avatar

Highlights

  • Pro

Block or report mself8

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
mself8/README.md

Songyun Jeong, Sports Analytics × Machine Learning. A dot-style recreation of Rooney's bicycle kick against Manchester City in 2011.

LinkedIn Email

About

I'm passionate about turning complex data into decisions people can trust, especially in sports. As an undergraduate researcher at the CIDA Lab, University of Seoul, I use soccer event data (tabular) and tracking data (time-series) to build models for lineup recommendation and offline evaluation, as well as diffusion models that generate player movement. I also develop LLM agents that answer from retrieved documents. My goal is to build AI products that prove their value through careful evaluation and make a real difference for teams, businesses, and users.

Tech Stack

Python PyTorch scikit-learn pandas NumPy PyG Jupyter LangGraph Hugging Face vLLM MCP PostgreSQL Streamlit Docker OpenCV D3

Publications & Talks

  • Recommending the Starting Eleven: Counterfactual Lineups from Event and Tracking Data
    First author · MIT Sloan Sports Analytics Conference 2027, Research Paper Competition (abstract accepted) · Code
  • Soccer Lineup Optimization based on Spatial Action Values
    First author · KCC 2026 · Code
  • Spatial Action Value Based Lineup Optimization in Soccer via Differentiable Player Selection
    Oral presentation · NU Sports ML Workshop, Nagoya University, 2026 · Code

Featured Projects

Set-Piece Danger Maps on Broadcast Video
Maps broadcast pixels to pitch coordinates with homography and overlays VAEP-based danger heatmaps on K League footage.
OpenCV · VAEP
Research Ethics Review Routing Agent
Routes a study plan to the right ethics review and cites the legal clauses behind each decision. 94% routing accuracy on a sealed set of 600 cases.
LangGraph · MCP · vLLM
Clinical Note Verification
Detects hallucinations in LLM-written clinical notes and links each sentence to its source in the dialogue. Caught 90.6% of 139 injected errors.
vLLM · LLM-as-a-judge
Wildfire Spread Simulator
Arrival-time cellular automaton trained on U.S. and Portuguese fires and applied to large Korean wildfires.
Cellular automata · Geospatial
Satellite Change Detection
Multi-sensor mapping of the Grindavík eruption lava field and the Incheon Airport expansion.
Sentinel-1/2 · Deep learning · Web map

Awards

  • 2026.10 Innovation Award (NST Chairman's Award) · NAIS AI Hackathon 2026
  • 2026.09 1st Place · AI Career Path Program
  • 2026.07 Excellence Award (Undergraduate Division) · KCC 2026
  • 2026.01 1st Place · K League × UOS Open AI Competition, Track 2 (DACON) · team lead
  • 2026.01 2nd Place · UOS Global Startup Camp

Pinned Loading

  1. Recommending-the-Starting-Eleven-Counterfactual-Lineups-from-Event-and-Tracking-Data Recommending-the-Starting-Eleven-Counterfactual-Lineups-from-Event-and-Tracking-Data Public

    Recommending the Starting Eleven: Counterfactual Lineups from Event and Tracking Data (SSAC 2027 abstract code)

    Python

  2. spatial-vaep-team-builder-v2 spatial-vaep-team-builder-v2 Public

    Spatial Action Value Based Lineup Optimization in Soccer via Differentiable Player Selection (CIKM 2026)

    Python 1

  3. irb-route-agent irb-route-agent Public

    NAIS AI 해커톤 2026 · 심의 경로 판정 AI 에이전트 (IRB·DRB 관할 판정, 조문 근거 추적, 판단불가 명시)

    Python

  4. k-track2 k-track2 Public

    VAEP-based set-piece danger heatmaps overlaid on K-League broadcast video via OpenCV homography, with an MVP dashboard — K-League data competition (Dacon) Track 2

    Jupyter Notebook

  5. soap-note-verification soap-note-verification Public

    진료대화→SOAP 노트 '검증 자동화' — 문장단위 근거추적([T#] citation)·LLM-judge 환각검출·함정셋으로 검증기 자체 성능 측정 (로컬 vLLM 다중모델 비교)

    Python