Building reliable scientific data systems and representation learning methods for computational drug discovery.
Academic snapshot
Computer Science undergraduate at Xidian University with research training across AIDD agents, ADMET data systems, molecular representation learning and peptide modeling.
Selected research
Current work centers on scientific data extraction, context-aware molecular modeling, and computational drug discovery.
At Tsinghua AIR ATOM Lab, building an agent-driven patent extraction and quality-control system for molecular structures, ADMET endpoints, assay conditions, values and units.
View related outputs→Designing a unified two-tower ADMET prediction framework that treats assay and experimental context as first-class model inputs rather than metadata left outside the model.
View model work→Built and trained a peptide instruction model with LoRA, covering function description, sequence design, property prediction and physicochemical optimization.
Open paper↗Contributed to 3D-MPG, a molecular representation learning manuscript on relational geometry pretraining through intramolecular region-pair arrangements, submitted to AAAI 2027.
View submission record→Built a task-adaptive molecular pretraining platform recognized with Xinghuo Cup university first prize and college-level special prize.
View award record→Exploring PharmAgent workflows that connect autonomous research planning, molecular evidence and downstream drug-discovery tasks.
View agent systems→Improving the sequence tower of a DrugCLIP-style protein-ligand foundation model with ESM-C representations and hyperbolic latent structure for OOD generalization across targets, scaffolds and low-data settings.
View ADMET work→Publications & outputs
Selected published and submitted work is presented with current status, authorship role and supporting records.
View published paper ↗Second author / undergraduate lead author. The work connects instruction-tuned language modeling with peptide engineering tasks and was published in Bioinformatics.
Second author. The manuscript was submitted to ICLR 2027 on 19 September 2026.
View submission record↗
Second author / undergraduate lead author. The manuscript focuses on 3D molecular geometry pretraining for molecular representation learning.
View submission record↗
Third author. The manuscript was submitted to ICLR 2027 on 19 September 2026.
View submission record↗
Third author. The manuscript builds a systematic benchmark for peptide property prediction.
View manuscript record↗
Seventh author. The manuscript was submitted to ICLR 2027 on 19 September 2026.
View submission record↗
The manuscript was submitted to the NeurIPS 2026 AI4DD Workshop on 6 September 2026.
View submission record↗
Life moments
Research training, academic sharing, international exchange and orchestra performance
Member of Xidian Symphony Orchestra, a Shaanxi high-level arts troupe and first-chair troupe of Shaanxi collegiate arts troupe; participated in major performances including the ICAI 2024 opening ceremony.
A college WeChat article presenting representative undergraduate work across research, academics, service and broader development
Click the card to switch between the front and back sides of the college publicity material
About
I care about one question:
I am a Computer Science undergraduate at Xidian University and a research intern at the Institute for AI Industry Research (AIR), Tsinghua University.
My current interests include AI for Drug Discovery, patent-scale scientific information extraction, high-quality ADMET data construction, and representation learning for molecules and proteins.
I am especially interested in systems that connect robust data infrastructure with predictive models that remain useful outside curated benchmarks.
Predictions should remain traceable to scientific evidence.
Model assay and experimental context, not only molecules.
Data quality is part of the scientific method, not cleanup.
Experience
Computer science → AI4Science → drug discovery.
AI for Drug Discovery, scientific data extraction, and autonomous research systems.
Computer science training with a research focus on AI for scientific discovery.
wutianxiang.cn