Ilgyu Yi

Machine Learning & Software Engineer · Seoul, Republic of Korea

Professional Summary

Engineer working across game AI, reinforcement learning, and decentralized systems, with a focus on simulation-based optimization, computational geometry, and combinatorial problem solving. Comfortable owning a system end to end — building the simulation environment an agent learns in, designing the learning architecture, and taking the result through to a running production service.

Work in Progress

Unirec — a modular, RL-friendly recommendation framework with swappable components, typed interfaces, and reproducible caching. github.com/ilgyu-yi/unirec

Professional Experience

Senior Engineer, Game AI Team — NHN

Republic of Korea · Apr 2026 – Present

Built handol-mux, a multiplexing engine for Handol — NHN’s AlphaGo Zero-based Go AI — tripling the number of concurrent games served on the same hardware with no loss of playing strength.
Developing successor engines: one that plays in a more human-like style, and one that exceeds the current engine’s playing strength.

Independent Professional Development

Republic of Korea · Jun 2025 – Mar 2026

Self-directed development period; consolidated and formalized a personal software development methodology.

Blockchain Engineer — Planetarium Labs

Republic of Korea, Singapore · Jun 2022 – May 2025

Maintained and extended Libplanet, an open-source blockchain framework for decentralized games.
Stabilized the PBFT consensus mechanism.
Designed and implemented core DPoS consensus mechanisms.
Implemented a VRF conforming to RFC 9381.
Implemented probabilistic sortition logic based on the Algorand paper.

Machine Learning Engineer — Spacewalk

Republic of Korea · Nov 2019 – Mar 2022

Main engineer on RL-based architectural design automation, implementing the pipeline from parameterization through optimization.
Built the training environment with computational geometry, turning land parcels and design constraints into the state and reward representations the agents acted on.
Designed autoregressive policies, pointer networks, and generalizable agents operating over dynamic state–action spaces.
Launched the Landbook Premium mobile app, generating a building design per land parcel in under three minutes.
Worked with the product team and practicing architects to evaluate design quality on an ongoing basis.

Machine Learning Engineer — Aidentify

Republic of Korea · Sep 2018 – Nov 2019

Developed an RL agent for HVAC optimization in data centers, combining mixture density networks, Dyna, and a modified value function to reach 30% energy savings in simulation.
Delivered ML research projects for external clients: domain adaptation and neural architecture search (ETRI), anomaly detection (Hyundai Autron), image segmentation (Ministry of Environment).

Supervisor, Raw Materials Team — Hyundai Steel

Republic of Korea · Jul 2017 – Sep 2017

Inspected and purchased scrap material, and established scrap input plans for the mill.

Project Engineer, Mechanical Engineering Team — DL E&C

Republic of Korea · Jan 2014 – Aug 2014

Handled scheduling, cost estimation, and expediting for mechanical instruments, piping, and manpower.

Education

Yonsei University — B.S. in Mechanical Engineering (2008 – 2014)

Technical Skills

Languages: Python, C#

Frameworks: PyTorch, TensorFlow, scikit-learn, NumPy, Pandas

Domains: Reinforcement Learning, Machine Learning, Computational Geometry, Probabilistic Modeling, Combinatorial Optimization, Simulation, Distributed Ledger, Open Source

Tools: Git, Docker, Rhino/Grasshopper, EnergyPlus, Unity

Patents & Open Source

Patents (Republic of Korea): KR 10-2698899, KR 10-2698858, KR 10-2118088, KR 10-1988504

Open Source: Contributor to blockchain infrastructure, primarily Libplanet.