Seongok Ryu

Seongok Ryu

Founding Research Scientist at GALUX, an AI-driven drug discovery startup in South Korea.

Background in Quantum Chemistry and Machine Learning with experience spanning small molecule to antibody discovery. Focus areas include Geometric Deep Learning, Generative Models, and Bayesian Deep Learning for accurate biomolecule description. Current goal: pioneering Agentic Virtual Drug Discovery through autonomous AI agents managing literature mining, multi-modal data analysis, de novo protein design, and large-scale screening.

Professional Experience

GALUX Inc. — Founding Research Scientist

Seoul, South Korea | April 2021 – Present

Joined as first employee, establishing core R&D infrastructure.

AI-driven Protein Structure Prediction & Design:

  • Enhanced hit rate of de novo protein binder discovery via Antigen-Antibody complex structure prediction
  • Developed template-based and hotspot-guided docking strategies
  • Engineered pipelines using UniFold and Boltz with memory-efficient training on RTX 4090 and H100 clusters
  • Implemented Latest Weight Averaging (LAWA), BioEmu augmentation, novel loss functions

Small Molecule Drug Discovery Platform:

  • Built end-to-end AI platform with Active Learning for HTVS and Fragment-Based Drug Design (FBDD)
  • Developed GNNs for ADME/T prediction and GalaxyDock2-DL for binding pose prediction

AITRICS — Team Lead / Research Scientist

Seoul, South Korea | February 2020 – March 2021

  • Led AI Drug Discovery team establishing predictive and generative modeling pipelines
  • Established benchmarks for Graph Neural Networks and Bayesian Learning on small datasets
  • Developed Reinforcement Learning-based molecular design algorithm optimizing docking scores
  • Published two papers as corresponding author in NeurIPS and JCIM

Kakao Brain — Research Intern

Seongnam, South Korea | December 2018 – February 2019

  • Developed molecular property prediction models using multi-task learning and GNNs
  • Implemented GNN-based models for protein-ligand binding affinity prediction

Technical Skills

Programming & Frameworks

  • Languages: Python, C/C++
  • AI Frameworks: PyTorch, PyTorch Geometric, Deep Graph Library (DGL)
  • Structure Prediction: AlphaFold, UniFold, Boltz

AI & Algorithms

  • Graph Neural Networks (GNN)
  • Geometric Deep Learning
  • Bayesian Deep Learning
  • Generative Models & Reinforcement Learning (RL)

Computational Chemistry

  • Structure-based Drug Design (SBDD)
  • Fragment-Based Drug Design (FBDD)
  • Molecular Docking & Dynamics
  • Quantum Chemistry (DFT)

HPC & Tools

  • Linux, Slurm, PBS, Snakemake

Education

KAIST — Ph.D. in Chemistry

March 2014 – February 2020

  • Thesis: “Development of machine learning systems for drug discovery”
  • Advisor: Prof. Woo Youn Kim
  • Shifted research post-AlphaGo toward applying GNNs, Bayesian Deep Learning, and Generative Models
  • Early phase work: contributed to ACE-Molecule (DFT software) development and implemented numerical geometry optimization (L-BFGS) in C++
  • Co-designed curriculum for “Artificial Intelligence and Chemistry” course

KAIST — B.S. in Chemistry and Physics (Double Major)

February 2009 – February 2014

  • Recipient: Korea Presidential Fellowship (144 nationwide recipients selected by South Korea’s President)

Contact