Live model
Score peptide–MHC binding and antigen presentation from sequence; batch scoring and benchmark uploads supported.
Open deepneo.kevinwanglab.org →We are witnessing an extraordinary historical inflection point: biology has ceased to be a purely descriptive discipline and has transformed into a programmable, predictive systems science. The rapid maturation of generative deep learning—from structural breakthroughs like AlphaFold 3 resolving multi-entity interactions, to foundational protein language models like ESM2 treating amino-acid sequences as structured linguistic tokens—has fundamentally rewritten the rules of therapeutic discovery. My drive to enter this field is fueled by a singular, intense realization: AI is no longer just an optimization tool for the wet lab; it is finally allowing us to decode the underlying grammar of the immune system and anticipate cellular behavior before an experiment even begins.
While my passion is driven by this new frontier of computational medicine, my baseline capability is anchored in a strong background in large-scale systems engineering and data architecture. In the current era of AI for Science, the primary bottleneck is no longer model size, but data engineering—aligning highly fragmented, multimodal biological data into structured, auditable streams that deep architectures can actually learn from. My engineering experience is not a separate past life; it is the exact technical prism through which I approach biological data pipelines, giving me the unique ability to build frameworks that are robust, reproducible, and mathematically sound.
This intellectual pull led me to return to university for my MSc at Auckland University of Technology (AUT), dedicating my research to cellular immunology, peptide–MHC presentation dynamics, and the computational design of neoantigen cancer vaccines. My thesis project, DeepNeo-CL, is my first concrete contribution to this field. Rather than creating an isolated, theoretical notebook, I developed a peptide–MHC predictor utilizing the ESM2 transformer framework. Crucially, I pushed this project into a fully deployed, benchmarked live computational architecture on cloud infrastructure—ensuring the model’s predictions are transparent, auditable, and ready for scientific scrutiny.