const yohan = {
role: "AI Research Engineer @ KakaoBank 🏦",
focus: ["Reasoning Systems", "Memory & Cognition", "Post-Training"],
scale: "26M+ Active Users 🚀",
mission: "Making LLMs Think Like Humans 🧬",
passion: [
"🔮 Parallel & Continuous Reasoning",
"⚡ Scalable LLM Optimization (200B+ params)",
"🧬 Episodic Memory & Cognitive Modeling",
"🛡️ Trustworthy & Aligned AI Systems"
]
};💡 I don't just scale parameters — I reimagine how models think, remember, and align with human cognition.
# Current LLMs: Sequential 🐌
for token in sequence:
output.append(generate(token))
# My Vision: Parallel & Continuous 🚀
thoughts = parallel_reasoning(context)
output = diffuse_and_refine(thoughts)Inspired by: Soft Token Reasoning Goal: Models that think before they speak |
@ KakaoBank: 200B+ Parameter LLMs 🎯 Key Innovations:
|
|
Building PREMem → Next-Gen Memory Systems graph LR
A[Experience] --> B[Consolidate]
B --> C[Structured Memory]
C --> D[Contextualize]
D --> E[Reasoning Loop]
E --> F[Decision]
Research: Episodic memory that learns to forget |
🎯 Core Principles:
Vision: AI that reasons with ethics, not just logic |
🔥 Featured Papers
📄 Finding Diamonds in Conversation Haystacks 💎
Yohan Lee, Yongwoo Song, Sangyeop Kim†
📄 Pre-Storage Reasoning for Episodic Memory 🧠
Sangyeop Kim*, Yohan Lee*, Sanghwa Kim, Hyunjong Kim, Sungzoon Cho†
PREMem: Shifting inference burden to memory for smarter dialogue
📄 What Really Matters in Many-Shot Attacks? 🛡️
Sangyeop Kim*, Yohan Lee*, Yongwoo Song*, Kimin Lee†
📄 HEISIR: Hierarchical Expansion of Inverted Semantic Indexing 🔍
Sangyeop Kim†, Hangyeul Lee, Yohan Lee
📄 SAFARI: Sample-specific Assessment Framework 📊
Yohan Lee*, Sungho Park*, Sangwoo Han*, Yunsung Lee*†, and team
🧠 LLM Post-Training | ⚡ Inference Optimization | 🔬 Reasoning Systems | 🧬 Memory Architectures | 🛡️ AI Safety
mindmap
root((Yohan's<br/>Research))
🧠 Reasoning
Parallel Inference
Soft Token Reasoning
Diffusion-like Dynamics
⚙️ Production
200B+ LLM Optimization
Financial AI Agents
Real-time Reasoning
🧬 Memory
PREMem Evolution
Episodic Systems
Contextual Learning
🛡️ Safety
Alignment Research
Trustworthy AI
Ethical Reasoning
