Simon (Seungwon) Lee

AI & Systems Researcher · Mathematical Foundations & Neuro-Symbolic Computing
Email: redcar1024@gmail.com GitHub: github.com/Heisnotanimposter LinkedIn: simon-lee-003689288 Location: Seoul, Republic of Korea

Focusing on neuro-symbolic artificial intelligence, discrete program synthesis under spatial constraints, and real-time state estimation for autonomous systems. Ranked in the Top 4% Globally in ARC Prize 2026 (Silver Medal, 194th / 4,795+) and Top 13% in ARC Prize 2025 (Abstraction and Reasoning Corpus). Developer of high-performance interactive scientific platforms, 3D spatial simulation engines, and rigorous open-source implementations of graduate machine learning curricula (Stanford CS229 & CS230).

Domain / Track Core Contributions & Milestones Reference / Artifact
ARC Prize 2026 Top 4% Global Ranking — Silver Medal (194th / 4,795+ teams, Score: 33.38). Verified Closed-Loop Control (VCLC) & Sub-Millisecond Verification Engine (RVICS). HINAGI 2026 Repo ↗
ARC Prize 2025 Top 13% Global Rank. Multi-dimensional neuro-symbolic program synthesis engine with discrete spatial primitives and cellular automata. ARC-AGI-HIN Repo ↗
Running Interactive Services Live WebGL 3D systems: Stanford CS229 Platform, SafeRoute Sense 3D spatial simulator, Protein Explorer (AlphaFold PAE). View Running Services ↓
Graduate Curricula & Proofs Rigorous mathematical solutions and derivations: PS0–PS4 (GLM, SVM, EM, MDP), Matrix Calculus, ZBDD reduction rules. Inspect Proofs & PS ↓
Curriculum Vitae 4-Language verified academic resume (English, Korean, Japanese, Chinese) for university admissions and institutional review. Read Full Academic CV ↓

I. Interactive Services & Core Research

Interactive web-deployed scientific platforms, open-source research implementations, and algorithmic systems. Use the search input or category filters to navigate.

1. Stanford CS229: Machine Learning Interactive Platform Running Service · Online
Andrew Ng's Stanford Graduate ML Curriculum · Interactive Visualizations & 3D Loss Surfaces

A comprehensive pedagogical web platform expanding Stanford's flagship CS229 curriculum. Implements real-time 3D loss surface exploration, dynamic decision boundary rendering for Support Vector Machines and Logistic Regression, and step-by-step mathematical derivations for Problem Sets PS0 through PS4.

Key Algorithms: Newton-Raphson, Support Vector Machines (Lagrange Multipliers, SMO), Gaussian Discriminant Analysis (GDA), Expectation-Maximization (GMM), Bellman Equations.
Technology Stack: WebGL, Three.js, PyTorch, MathJax, TypeScript, GitHub Pages.
[ Launch Running Service ↗ ] [ GitHub Repository ↗ ]
2. SafeRoute Sense 3D: Autonomous Spatial Navigation Engine Running Service · Online
Real-Time 3D Spatial Risk Simulator · Dynamic Collision Cone Routing

WebGL-based autonomous spatial navigation simulator evaluating dynamic pedestrian safety and hazard avoidance in urban environments. Features dynamic collision cone computations, multi-agent trajectory prediction, and synthetic LiDAR ray-casting in simulated 3D city terrain.

Key Algorithms: Dynamic Collision Cone (DCC), Dijkstra/A* Spatial Search, Kinematic Vehicle Models, LiDAR Ray-Marching.
Technology Stack: WebGL, Three.js, Vite, TypeScript, Canvas API.
3. Protein Explorer & AI Assistant (AlphaFold PAE) Running Service · Online
Molecular Structure Visualizer · Predicted Aligned Error Heatmaps

Interactive 3D biological macromolecule visualizer rendering PDB structures directly in the browser via 3Dmol.js with an automatic RCSB PDB CloudFront CDN fallback. Features Predicted Aligned Error (PAE) interactive heatmaps, residue distance matrices, and secondary structure ribbons.

Key Algorithms: 3D Coordinate Transformation, PDB Parser, PAE Matrix Heatmap, Residue Distance Calculation.
Technology Stack: 3Dmol.js, JavaScript (ES6+), RCSB CloudFront CDN, HTML5.
4. ARC-AGI-HIN: Multi-Dimensional Neuro-Symbolic Program Synthesis Core Research · Top 13%
Top 13% Global Rank in ARC Prize 2025 · Discrete Logic Primitives & Cellular Automata

End-to-end neuro-symbolic program synthesis engine developed for the ARC Prize 2025 (Abstraction and Reasoning Corpus). Features a domain-specific language (DSL) with 120+ spatial transformation primitives, bidirectional state-space search, and ZBDD-compressed hypothesis spaces.

Key Algorithms: Domain-Specific Language (DSL) Synthesis, Bidirectional Breadth-First Search, Cellular Automata, Invariant Feature Extraction.
Technology Stack: Python, PyTorch, NumPy, Multiprocessing, Rust bindings.
[ GitHub Repository ↗ ]
5. Probabilistic-State-Estimation: Real-Time Sensor Fusion & Filtering Core Research
Bayesian Filtering Suite · Extended Kalman Filter (EKF) & Unscented Kalman Filter (UKF)

High-performance Bayesian filtering and state estimation suite designed for non-linear kinematic bicycle models and noisy IMU/GPS telemetry. Evaluates non-linear state propagation, unscented transform sigma point generation, and particle filtering under non-Gaussian noise distributions.

Key Algorithms: Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), Particle Filter (Sequential Monte Carlo), MSSA Missing Value Imputation.
Technology Stack: C++, Python, NumPy, SciPy, Matplotlib.
6. AutoML Ops: Automated Model Selection & Hyperparameter Optimization Systems & Engineering
Automated Machine Learning Pipeline · Distributed Hyperparameter Tuning

Cloud-ready automated ML engineering system providing end-to-end dataset preprocessing, Bayesian hyperparameter optimization (Optuna / TPE), automated feature engineering, and cross-validated ensemble stacking.

Key Algorithms: Tree-structured Parzen Estimator (TPE), Genetic Feature Selection, Stacking Ensembles, Early Stopping.
Technology Stack: Python, Scikit-Learn, LightGBM, XGBoost, Optuna, FastAPI.
7. MarketPulse: Real-Time High-Frequency Quantitative Telemetry Systems & Engineering
Order-Book Microstructure Analytics · Streaming Market Signals

High-throughput telemetry pipeline aggregating order flow imbalance, microstructural volatility, and depth distribution across cryptocurrency and financial exchanges in real time via asynchronous WebSockets.

Key Algorithms: Order Flow Imbalance (OFI), Volume-Weighted Average Price (VWAP), Real-Time Sliding Window Statistics.
Technology Stack: Python, Asyncio, WebSockets, Pandas, Redis.
Graduate Curricula & Academic Records
Inspect mathematical problem sets and proofs or review the full academic curriculum vitae.

Graduate Curriculum & Mathematical Proofs

Detailed mathematical solutions, closed-form derivations, and algorithmic implementations for graduate coursework and foundational machine learning theories. Click any module to inspect step-by-step proofs and derivations.

1. Stanford CS229: Machine Learning
Instructor: Prof. Andrew Ng · Stanford Graduate School · Autumn 2018

Foundational Scope: Supervised Learning (Linear Regression, Logistic Regression, GLMs, GDA, Naive Bayes), Support Vector Machines (Lagrange Multipliers, Dual Formulation, Mercer's Kernels, SMO), Learning Theory (PAC Learning, VC Dimension), Unsupervised Learning (K-Means, EM Algorithm for GMMs, PCA, ICA), Reinforcement Learning (MDPs, Value Iteration, Policy Iteration, Bellman Equations).

[ Open CS229 Interactive Platform ↗ ]
2. Stanford CS230: Deep Learning
Instructor: Prof. Andrew Ng · Stanford University

Foundational Scope: Vectorized Matrix Calculus Formulation of Backpropagation, Tensor Shape Invariance, Deep Feedforward Networks, Optimization (SGD with Momentum, RMSprop, Adam Bias Correction), Regularization (L2, Dropout), Batch Normalization Forward/Backward Equations, Convolutional & Sequence Architectures.

3. MITx 6.431x / 18.6501x: Fundamentals of Statistics
MITx MicroMasters in Statistics and Data Science

Foundational Scope: Convergence of Random Variables (Weak/Strong Law of Large Numbers, Central Limit Theorem), Maximum Likelihood Estimation (MLE), Fisher Information Matrix, Cramér-Rao Lower Bound, Confidence Intervals, Hypothesis Testing (Neyman-Pearson Lemma, Generalized Likelihood Ratio Test).

4. Zero-Suppressed Binary Decision Diagrams (ZBDD) & Discrete Logic
Mathematical Foundations: Donald Knuth (TAOCP Vol 4A) & Shin-ichi Minato (1993)

Foundational Scope: Canonical Graph Representations of Combinatorial Families, Elimination Rule for 0-edges (ZBDD) vs Node Elimination (standard BDD), Shannon Decomposition, Polynomial-Time Set Union/Intersection Primitives, Memoized Recursive Path Counting for Constraint Satisfaction.

Navigation
Return to Showcase & Projects or review the full Academic CV.

Simon Lee

Email: redcar1024@gmail.com | Phone: +82-10-7726-2102 | Line ID: ngc1024 | Kakao ID: redcar359 | WeChat ID: wxid_acxnq69n3f1912 | Location: South Korea
GitHub: github.com/OpenOurMinds | LinkedIn: linkedin.com/in/simon-lee-003689288 | HuggingFace: huggingface.co/ngc1024

Research Interests

Education

Academic Admissions & Recognition

Research & Projects

Running Services (Active Apps)

Technical Skills

Certifications & Military Service