About Me

01

Hi, I'm Kehan, a final-year undergraduate in Information and Computing Science at the School of Mathematics and Statistics, Central South University. My undergraduate work lies at the intersection of computational mathematics and geophysics.

My current research focuses on medical large language models, virtual cells, AI for Science, and benchmarking. My experience includes evidence-based medical agents, systematic evaluation of single-cell models, and high-performance numerical simulation.

My next destination is the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, where I will join Prof. Min Yang's group.

Medical LLMsVirtual CellsAI for ScienceBenchmarking
Let's connect

I welcome conversations about medical AI, virtual cell models, and reliable model evaluation.

Get in touch

Research Experience

02
Research internship · SIAT, CAS

Evidence-based Medical Agents

Supervised by Prof. Min Yang

Contributed to a two-stage agentic workflow for evidence-based diagnosis and dynamic treatment, with attention to verifiable evidence and information transfer between stages.

  • Helped build a 5A workflow: Ask, Acquire, Appraise, Apply, and Audit, with agent loops for checking and revisiting intermediate steps.
  • Contributed to supporting methods in the reasoning chain, frontend interface design, and server deployment.
Medical LLMsAgentic WorkflowEvidence-based AI
Research internship · SIAT, CAS

Systematic Benchmarking of Single-cell Models

Supervised by Prof. Min Yang

Participated in a unified evaluation framework for AI virtual cell models, organized around three core capabilities and eight task categories.

  • Reproduced model tutorials and integrated models including scGPT, PertAdapt, CellFM, and CAPTAIN into a common evaluation workflow.
  • Worked on gene perturbation and cell-type annotation tasks using datasets such as Adamson, Norman, and Dixit, with shared metrics and configuration-based execution.
AI for ScienceSingle-cell ModelsBenchmarking
Research project · Central South University

High-performance 3D Seismic Wave Simulation

Supervised by Prof. Kejia Pan

Studied frequency-domain spectral-element modeling of seismic waves in viscoelastic anisotropic media with irregular topography, focusing on sparse linear-system solvers and parallel efficiency.

  • Compared MUMPS, AGMG, BoomerAMG, and AMGCL; evaluated preconditioning, numerical libraries, and hybrid OpenMP/MPI configurations.
  • On the evaluated configurations, AGMG reduced serial runtime by 91.59% relative to MUMPS; the tuned parallel setup achieved approximately 13× speedup with over 90% lower memory use.
Scientific ComputingFortranMPI / OpenMP

Education

03

Central South University

Final-year Undergraduate

School of Mathematics and Statistics

Information and Computing Science · Computational Mathematics × Geophysics

91.13/100Weighted Average
3/82Academic Ranking
1/82Comprehensive Ranking

Academic record as documented in my 2026 résumé.

Selected coursework: Operations Research (100), Data Structures (97), C++ Programming (97), Numerical Computing Practice (97), Digital Signal Processing (96), Mathematical Analysis III (95).

Honors & Awards

04
  • National
    National ScholarshipAwarded during the second undergraduate year
  • National
    China Undergraduate Mathematical Contest in Modeling · Second PrizeTeam leader · modeling, paper writing, and supporting code
  • Provincial
    National College Student Market Survey and Analysis Competition · First Prize, HunanData analysis, programming, and model development
  • Provincial
    Chinese Mathematics Competitions · First Prize
  • Provincial
    National College Student Statistical Modeling Competition · Third PrizeNLP · sentiment modeling with an improved BERT-based model

Technical Background

05
Languages
Python, C++, Fortran, MATLAB
Tools & Methods
Git, Linux, SSH, LaTeX, MPI / OpenMP, model fine-tuning & evaluation