Coding is my primary tool for creativity — transforming ideas into tangible software and applications. Beyond research, I develop side projects that let me rapidly prototype solutions to real-world needs and use them as small-scale testbeds for studying human–AI interaction, retrieval-augmented reasoning, and practical evaluation methods. These projects leverage my work in NLP, knowledge graphs, and RAG pipelines, and they help sharpen skills in prompt engineering, context management, and full-stack implementation. They also motivate my PhD aspiration to deepen the theoretical foundations of trustworthy, human-centered AI.
These side projects demonstrate my drive and ability to translate research ideas into working prototypes and practical solutions — a research→application workflow I have followed from academic work to industry (e.g., GraphRAG). They serve as focused testbeds to explore design choices, prompt engineering, and interaction patterns that inform more rigorous research questions about reliability, interpretability, and human-centered AI — aligning with interdisciplinary and application-oriented research goals.
Mobile App — Shared diary & anniversary assistant (MVP)
This project began from a personal use case and evolved into a working MVP within a short timeframe.
React Native (Expo) · Supabase · LLM API · MVP · Expected release: Dec 2025
A mobile app for partners or individuals to manage shared anniversaries and personal diaries. Integrates a context-aware assistant that proposes personalized replies and reminders informed by the stored context.
- Diaries & Anniversaries storage (private or partner-shared, offline mode and online mode both accessible)
- Personalized reminders with AI-generated suggestions based on user context and notifications.
- Context-aware AI assistant for personalized interactions
- Data storage and backup via Supabase
Screenshots and demos will be provided after the internal beta and MVP release by the end of December 2025.
Web Platform — Knowledge management & AI-assisted writing (Prototype)
A prototype platform applying RAG and KG concepts to support knowledge extraction and evidence-backed writing assistance.
Next.js · FastAPI · PostgreSQL · ChromaDB · Prototype
An LLM-driven platform combining knowledge digestion and content creation, aiming to explore human-aligned, adaptive AI behaviors in knowledge-intensive tasks. Users can upload documents, automatically extract structured knowledge, and receive context-aware, evidence-backed suggestions to support reasoning and writing.
- Research / testbed role: Intended as a research-oriented prototype to study retrieval-augmented reasoning, human-in-the-loop adaptation, safe LLM generation, and interactive evaluation protocols.
- Planned capabilities: document ingestion → lightweight KB construction → evidence-backed suggestions for drafting and revision (work-in-progress).
-
Planned research use:
- Evaluate factuality, interpretability, and alignment with user intent
- Collect lightweight user feedback to explore adaptive generation and interactive reasoning
- Develop and test small-scale methods for auditing, evidence presentation, and safe AI adaptation.
Status: in progress(very early prototype, under development.)
Prototype screenshots and demos can be provided after stabilization. Detailed demonstrations can also be arranged during potential academic meetings.
Why These Projects Matter for Research
These side projects are intentionally small-scale and iterative, serving as early prototypes rather than finished products. They provide concrete contexts in which to test ideas about human-aligned reasoning, adaptive interaction, and reliability in LLM-augmented systems — research questions I plan to study more formally in a PhD. They allow me to experiment with human-in-the-loop adaptation, design evaluation protocols for factuality and alignment, and validate the feasibility of interactive AI behaviors before extending them into rigorous theoretical and empirical studies.