Liwei Li

Leveraging advanced AI models and AI integrated computational tools to accelerate design, optimization, and translation of biologics-based therapeutic candidates

Led the transformation of the company’s computational platform from traditional physics-based modeling to AI- and ML-driven design. Directed development of deep learning models that significantly improved prediction accuracy, design success rate, and turnaround time. Developed and fine-tuned machine learning and protein language models (ESM2) models for protein stability, solubility, and activity optimization. Applied advanced tools such as AlphaFold2, Boltz-2, RFdiffusion, ProteinMPNN, and ESM-IF1 for protein-small molecule structure prediction and protein backbone and sequence design. Deployed in-house GPU-powered molecular dynamics simulations and leveraged AWS cloud infrastructure (EC2, S3) to scale computational modeling, streamline data storage, and accelerate large-scale AI model training. Combined strategic leadership with hands-on modeling to build scalable AI infrastructure, enabling faster design–build–test–learn cycles and improving success rates across discovery programs.

Hands-on experience in antibody discovery and optimization, with a focus on improving affinity, stability, solubility, and overall developability. Experienced in assessing and mitigating manufacturability and immunogenicity risks through structure-based modeling and biophysical property analysis. Extensive background in small-molecule virtual screening, hit-to-lead optimization, and compound prioritization using docking, FEP, and cheminformatics workflows. Integrates machine learning and predictive modeling to support ADME property prediction and guide data-driven compound selection. Collaborates closely with cross-functional experimental teams to translate computational insights into validated design outcomes, accelerating discovery and improving success rates.

Applied computational protein design to improve enzyme specificity, stability, and solubility. Developed homology models, performed docking and molecular dynamics simulations, and guided variant design using bioinformatics insights. Designed and Optimized gene libraries to enhance expression and balance metabolic pathways. Collaborated with experimental teams to test and validate computational designs through molecular assays.

Developed machine learning–based methods for small-molecule discovery and design, leading to the identification of novel compounds targeting cancer-associated proteins such as GTPase Ral, urokinase receptor, and tissue transglutaminase. Performed large-scale virtual screening of millions of compounds and extensive molecular dynamics simulations to evaluate binding mechanisms and optimize lead candidates. Utilized RDKit for cheminformatics analysis, property prediction, and data curation to support structure- and ligand-based optimization. Published high-impact research in Nature and JACS, demonstrating advanced computational methodologies that improved accuracy and efficiency in small-molecule design.

Focused on physics-based force field development and molecular dynamics (MD) simulations of polymers and biological membranes to study molecular interactions and material behavior. Performed simulation-based free energy calculations to quantify thermodynamic properties and molecular interaction energies. Completed graduate coursework in Quantum Mechanics, strengthening the theoretical foundation for force field development and molecular modeling.