Computational biology · Single-cell perturbation modeling · Translational bioengineering
Tsinghua University · Beijing, China
My training spans Chemical Biology and Biomedical Engineering at Tsinghua University (GPA 3.98/4.00; rank 1/15). I work at the intersection of single-cell omics, perturbation modeling, gene-regulatory networks, and translational bioengineering.
I am especially interested in AI virtual cells that lead to actionable biological hypotheses. In practice, that means caring as much about leakage-aware evaluation, sample-level inference, reproducibility, and alternative explanations as I do about the model itself. My wet-lab background in biomaterials, molecular diagnostics, and neurobiology helps me keep computational questions connected to experiments.
- Single-cell perturbation and virtual cells — reproducing and extending perturbation-prediction models, representing genetic, chemical, and protein conditions, and testing whether performance transfers across datasets.
- Gene regulation and cell fate — reconstructing regulatory programs in megakaryocyte maturation and studying cell-state transitions in cancer persistence, hematopoiesis, and early embryonic development.
- Evidence-first computational biology — pseudobulk inference, trajectory and regulon analysis, cross-dataset validation, frozen analysis contracts, and adversarial checks against overinterpretation.
- Translational bioengineering — connecting omics-derived hypotheses with FACS, hydrogel biomaterials, neural interfaces, molecular diagnostics, and experimental validation.
A local-first desktop workbench for human-assisted LIF–MS annotation review. It combines synchronized signal and UMAP views, project-scoped SQLite state, reproducible preprocessing, and audit-friendly exports for Windows and macOS.
Python · SQLite · desktop software · scientific data review
A structured reproduction of X-Pert with native protein perturbation support, ESM-2 condition embeddings, stricter delta/DEG/pathway evaluation, and external validation on human immune single-cell datasets. The repository documents both what the extension can recover and where its claims should remain conservative.
Python · PyTorch · single-cell perturbation · AI4Science
A parallel agent-based simulator for embryonic lineage growth, time-windowed LoxCode barcoding, probabilistic tissue-fate assignment, and downstream fate-coupling analysis.
Python · Numba · R · lineage tracing
A privacy-first desktop language-learning app with AI-assisted content extraction, SM-2 spaced repetition, local SQLite storage, document import, and Windows/macOS releases.
Tauri · JavaScript · Rust · SQLite
Away from research, I am usually listening to ReoNa or turning a small idea into a working tool. Recent experiments include an evidence-grounded biology study assistant, a German speaking-practice companion, personal literature and knowledge agents, and a ReoNa-inspired desktop companion. I also enjoy building smaller projects such as Daily Star Gift Studio and PolyLingo.
I like agents that do more than produce a plausible answer: they should preserve sources, expose uncertainty, keep an audit trail, and leave a human with the final decision.
Computational biology: single-cell RNA-seq, pseudobulk differential expression, functional enrichment, gene-regulatory networks, regulon interpretation, trajectory analysis, cellular perturbation modeling, ESM-2, RDKit, and inverse-signature matching.
Programming and product engineering: Python, R, MATLAB, C, Git/GitHub, FastAPI, React, Tauri, SQLite, reproducible pipelines, desktop packaging, and scientific visualization.
Experimental: PCR, LAMP, Western blot, FACS, fluorescence microscopy, hydrogel synthesis and characterization, SEM/TEM, FTIR, SAXS, WXRD, DSC, microfluidics, and mouse stereotaxic workflows.
- Academic Excellence Scholarship, Tsinghua University — 2023, 2024, and 2025
- Comprehensive Excellence Scholarship, Tsinghua University — 2023
- First Prize, 39th National University Physics Competition (Non-Physics Group A) — 2023
I am happy to talk about single-cell perturbation modeling, virtual cells, scientific software, reproducible analysis, or an interesting small project that should exist.