Genomic sequences are modeled as evolving along binary phylogenetic trees through stochastic string-valued substitution and insertion-deletion (indel) processes. Given a set of present-day sequences, classical inference problems in phylogenetics are: (i) homology inference (ii) tree inference, and (iii) ancestral sequence reconstruction. A central focus of our recent work has been to develop fast frequentist indel-aware approaches to these problems. 

For tractability, the models in most cases must assume that residues evolve independently across sites. In reality, mutation probabilities are influenced by sequence context, including position-specific structural and functional constraints. In recent years, the convergence of computational biology and data-driven methods has led to genomic large language models (gLLMs). These can model sequence context dependences. 

Building on our previous work, our aim is to develop neuro-symbolic methods that retain mechanistic grounding of classical phylogenetics, and that integrate the representational richness of gLLMs. As a PhD student you will devise mutation models, develop inference algorithms, implement them in our Rust code-base, and evaluate the methods by simulation and on real data.

You should have a MSc in Computer Science, Computational Science, Computational Biology, Statistics / Applied Mathematics, or a related quantitative field, with a strong background in: 

Knowledge of phylogenetics, and/or an understanding of neural networks is an advantage.