An IP portfolio that proves a proprietary AI stack

Korean-specialized RLHF and Direct Preference Optimization (DPO) training pipeline

Context compression and retrieval-based memory management for multi-turn LLM dialogue

Explainable AI (XAI) system for conversational response generation with source attribution

Korean domain-specific RAG pipeline with integrated chunking, re-ranking, and hallucination detection

Intent-driven automatic prompt optimization and rewriting system for conversational AI

Automated quality evaluation and multi-dimensional refinement pipeline for AI dialogue training data