Undergraduate at National Chengchi University building language-model systems and exploring model behavior, evaluation, memory, and agentic architectures.
A stateful conversational AI system designed around continuity across sessions. Anjo combines semantic and emotional memory, persistent personality state, appraisal, and post-session reflection instead of treating each conversation as an isolated chat.
Stores semantic and emotional representations separately, then re-ranks retrieved memories using emotional context.
Models personality and mood using OCEAN and PAD-inspired state that persists and evolves across user sessions.
Post-session model analysis updates relationship state, emotional residue, trust, and personality deltas with bounded changes.
Designed and implemented a persistent-memory conversational system spanning backend architecture, retrieval, model orchestration, product behavior, and mobile delivery.
Worked on model-response evaluation and annotation tasks, including generalist and coding-oriented evaluation workflows.
Undergraduate in Communication, Taiwan. Independently building technical depth in machine learning, deep learning, language models, and AI systems.