At Peer Research Lab, I developed a dual-phase GraphRAG architecture combining Event Evolution Graphs with cognitive reasoning traces for multi-hop legal precedent analysis. My research there has also covered computer vision, NLP stance detection, explainable AI, and retrieval-augmented generation.
Through my deepfake detection thesis, I developed a dual-stream EfficientNet-B4 model that fuses spatial and FFT-frequency features and studied cross-dataset generalization. I also founded ARI Software Lab, where I published five mobile apps on Google Play with more than 1,000 combined downloads.

