CV Download PDF
AI4Bio · bioinformatics · machine learning. Click the button to download the full CV (PDF).
Contact Information
| Name | Chunzhuo Zhang (张淳卓) |
| Profession | AI4Bio |
| zhangchunzhuo@aaib.pku.edu.cn |
Experience
-
2024 - Biological Foundation Model Design
Research Institute
Designing single-cell foundation models and exploring multi-modal fusion for virtual cell modeling.
- Single-cell foundation model (53M / 300M / 600M / 2B): in-context learning with observation and perturbation data; group-level cell modeling. From data preprocessing through architecture design to downstream benchmarking (based on scGPT); outperforms scGPT, Transcriptformer, and UCE.
- Explored scaling laws on single-cell data.
- Implementation in PyTorch → PyTorch Lightning; attention acceleration via FlashAttention / FlexAttention; custom loss and multi-classification heads.
- Built a multi-task downstream evaluation pipeline (batch-effect correction, cell classification, perturbation prediction) and designed new tasks for single-cell foundation models.
- Distributed training: FSDP, data parallelism, activation checkpointing.
- Mentored interns to deliver independent projects.
- BindCraft protein design: designed antigen ligands with constrained binding sites and generated binders in collaboration with the experimental team; ~5% of designs showed strong binding affinity.
Education
-
2022 - 2024 Wageningen, Netherlands
M.Sc.
Wageningen University & Research
Bioinformatics
- Computational Biology (8.5/10), Deep Learning (8/10), Genomics (8/10), Advanced Statistics (8/10), Machine Learning (7.5/10), Molecular Systems Biology (7.5/10), Software Engineering (7.5/10)
- Protein structure & sequence embeddings (sgBERT, gBERT): combined protein structure and sequence via attention to obtain protein embeddings; preliminary exploration of multi-modal fusion. ChunZhuo/sgBERT
- Protein–protein interface false-positive prediction: extended DeepRank-GNN with an equivariant graph neural network using harmonic functions; benchmarked against HADDOCK, classical ML, and GNN baselines on the alanine-scanning task. ChunZhuo/SEGIN
- Coursework: sequence assembly, differential gene expression, protein domain analysis, gene regulatory network analysis; Java software development (parking-lot entry system).
-
2018 - 2022 Yangling, China
B.Sc.
Northwest A&F University (NWAFU)
Horticulture
- Advanced Mathematics (91/100), Linear Algebra (94/100), Probability Theory (92/100), Fundamental Biochemistry (91/100), Organic Chemistry (94/100)
- Kiwifruit shoot-tip cryotherapy for virus elimination (RT-PCR, plant tissue culture).
Skills
Programming: Python (PyTorch, PyTorch Lightning), R, Java, C++
ML Systems: FlashAttention, FlexAttention, FSDP, data parallelism, activation checkpointing
Domains: single-cell foundation models, protein representation learning, equivariant GNNs, multi-modal fusion
Languages
English : Fluent — IELTS 7.0 (Speaking 7.0); GRE 320
Mandarin Chinese : Native