Current approaches to genetic delivery remain largely empirical. Lipid nanoparticles carrying DNA, RNA, or antisense oligonucleotides are still optimised through iterative experimentation, with limited ability to predict their behaviour in vivo. Even with considerable improvement in organ-level targeting, there lacks a general framework for designing delivery to specific cell populations or reliably safety and immune activation.

Developing a machine learning framework for designing innate-immune-targeted lipid nanoparticles for the delivery of DNA, antisense oligonucleotides, and other genetic cargos. By combining de-novo protein design with predictive models of protein-corona formation and molecular binding, the framework will predict delivery, safety, and immune activation while engineering nanoparticles that selectively target defined innate immune cell populations.

By combining machine learning, de novo protein design, and the Stevens Group’s experimental delivery capabilities, the research project establishes a general framework for predictive genetic delivery The resulting approach will support the rational design of targeted delivery systems across multiple cargos, immune cell populations, and nanoparticle platforms.

Harsh recently completed his Master's in Molecular Bioengineering at Imperial College London, working at the intersection of machine learning and structural biology. In collaboration with Softnanolab, he is developing a protein language model specialised for predicting protein-protein interaction interfaces. Earlier, at the MRC Laboratory of Medical Sciences, he built a pipeline to retarget large protein toxins toward new receptor homologs. As an Encode fellow, he is bringing protein design into computational nanoparticle engineering with Professor Molly Stevens' group at the Kavli Institute, Oxford.