Yongji Lyu (吕勇稷)

About Me

Hello! I am Yongji Lyu (吕勇稷), an undergraduate student in the School of Engineering at Westlake University, majoring in Artificial Intelligence.

My research interests span deep learning, multimodal learning, and graph neural networks, with a particular focus on their applications in computational biology — especially spatial multi-omics data fusion and clustering. I am also actively exploring diffusion language models and multi-agent orchestration systems as a member of the Westlake University AGI Lab.

Download my CV

Education

Westlake University

B.Eng. in Artificial Intelligence, School of Engineering

Relevant coursework: Linear Algebra, Probability & Statistics, University Physics, Programming (C / Python), Machine Learning, Deep Learning, Data Structures & Algorithms

2025 — Present

Experience

Westlake University AGI Lab

Research Study Member

Focus: Diffusion language models (D3PM, SEDD), multimodal GNNs, contrastive learning (CLIP), DDPM — systematic study from mathematical derivation to code reproduction.

Survey: Multi-omics data integration (RNA / Protein / Peak) with graph attention networks; federated learning & privacy-preserving computing for GWAS.

2025 — Present

Westlake University Pebble Summer School

Teaching Assistant

Lecturer: "Machine Learning Applications in Biology" — introduced AI in genomics and protein structure prediction to PhD students in pharmacy and biology.

Designed interactive teaching content with visual materials to help students grasp complex ML concepts.

Jul 26 — Aug 5, 2026

Projects

STAMP: Spatial Multi-Omics Fusion Algorithm — A graph-neural-network-based method for spatial multi-omics (RNA + ATAC-seq) fusion and clustering, featuring a dual-graph GAT encoder with shared weights, asymmetric cross-modal attention (RNA→ATAC only), and cross-modal reconstruction losses. Achieved an average ARI of 0.903 on 5 simulated datasets (ranked 3rd among 10 methods) and validated on real mouse brain data (P22, ARI = 0.504 vs. cortical-layer annotations).

Knowpath Agent Orchestration — A multi-agent orchestration system that coordinates multiple AI agents on complex tasks, featuring centralized task dispatch, context passing, exception handling, and an auditable execution layer for traceability and interpretability.

Skills

Programming — Python (NumPy, Pandas, Matplotlib, PyTorch, Scikit-learn)

Machine Learning & Deep Learning — CNN, RNN, Transformer, Attention; diffusion models (DDPM, D3PM, SEDD), contrastive learning (CLIP), graph neural networks (GAT, GCN)

Cross-disciplinary — ML applications in biology (genomics, protein structure prediction)

Tools — Git, Linux CLI, Docker, ComfyUI, LaTeX, Jupyter Notebook, ArXiv, Papers with Code

English — IELTS (in preparation); fluent reading of academic papers and technical documentation; fully English-taught curriculum at Westlake University