Selected project pages
Projects are presented with concise summaries and technical tags, with a few pages expanded further when useful.
RepoLens
An evidence-grounded repository explorer that reconstructs architecture, execution flows, and risks from source code, with line-level citations and exportable reports.
Listening or Reading? An Empirical Study of Modality Importance Analysis Across AQA Question Types
A research case study investigating how audio QA systems trade off acoustic evidence against semantic and text-derived shortcuts across question types.
ECHOTWIN-QA: A Dual-Tower BEATSBERT System for DCASE 2025 Task 5 Audio Question Answering
A full end-to-end AQA system built for the DCASE 2025 challenge, including training, evaluation, ablations, and technical reporting.
ADAPTF-SEPNET: AudioSet-Driven Adaptive Pre-training of TF-SEPNet for Multi-device Acoustic Scene Classification
A collaborative challenge project focused on model development, experimental evaluation, and analysis for acoustic scene classification.
EmoSound: A Multimodal AI Agent for Emotion-Aware Audio Accompaniment of Emoticons
A multimodal AI agent project where I implemented components for the agent and evaluation pipeline, and contributed to experiments and writing.