Pejvak Moghimi

• Computational/ML lead of the Prenatal Skin Cell Atlas project within the Human Cell Atlas global consortium for 3D mapping of human body at single cell resolution.
• Discovered the signalling & gene regulatory network for angiogenesis in Prenatal skin and the top-25 relevant genes.
• Troubleshot, refactored and reproduced the results of >10,000 lines of code in 3 weeks to generate 50 publication-ready figures.
• Analysed large spatial and single-cell genomics data.
  Made significant contributions the research and the article being submitted to Science.
• Lead contributor to team's GitHub repository; creating clean, reproducible and beautifully documented GitHub repository, with ML-based interactive graphical mapping of this huge project.
• Established groundwork for reproducible data science.
• Updated department leads on research-based findings and provided any shortfalls or advances of note on weekly basis.
• Collaborated with biology domain experts to advance research and gain deeper understanding of topics.
• Reviewed and analysed historical and emerging literature to develop and practice new technologies.
• Contributed to and influenced discussions, seminars and lectures at Human Cell Atlas projects' events.

• Awarded ISMB fellowship for potential to continue delivering high-impact research in machine learning following successful demonstration of PhD results.
• Developed transformer-based LLM-type models of immune receptor molecules to augment NLP-based deep noisy long-tailed regression modelling of population-wide antibody commonality.
• Implemented self-supervised and transfer learning approaches for improving downstream supervised deep regression modelling.
  Led 4, and actively contributed to 1, open-source GitHub projects.
• Optimized previously published algorithms, such as Sumrep, and developed highly efficient graph-theoretical Python package capable of computing 100s of millions of sequences on personal laptops.
• Adhered to responsible-AI principles for creating model ready for clinical use.
• Achieved the most granular model of antibody commonality, as well as state-of-the-art predictive performance, by my deep conformal regressor model which quantifies epistemic and aleatoric uncertainty.

• Tutored 30 students on 3 modules of the “Bioinformatics with Systems Biology” MSc degree.
• Statistics: tutored 13 practical sessions on the highlighted topics of resampling and subsampling techniques, statistics with R, and unsupervised ML.
• Genomics: tutored 13 practical sessions on the highlighted topics of D&C algorithms, dynamical programming and Hidden Markov Models.
• BioComputing with Python: tutored 13 practical sessions on many computer science and Python programming topics.

Upon the success of my continuous computational genomics research in the group, I was hired by Prof. Thomas to continue my work to develop the function-discovery pipeline further (see Integrated Masters - research). I expanded the Python-based pipeline I had developed for querying databases, which helped with exploring the genomes of gram-positive bacteria to discover exporters of novel natural antibiotic compounds, which could be used as potential drug targets.

My first introduction to Python programming software development and data science. I completed multiple online courses on these topics from world-class universities, such as MIT and Harvard, while receiving hands-on training and supervision from computer scientists senior researchers in my group as well as the head of Genomics department.

I automated my phylogenomics pipeline - developed during my previous internship - through Python programming, which allowed me to access remote databases more readily and extend the power of the pipeline to a much larger and broader datasets. Notably, as part of an inter-departmental collaboration, I used this pipeline for analysing ABC transformers, among other more challenging transporter families, which resulted in successful identification of a broader range drug-targets.

First introduction to computer science, algorithms and computational biology and genomics, frequentist Vs. Bayesian statistical methods, statistical learning and mathematical inference.

Particularly focused on learning the foundations of the following within the context of the mathematical and statistical underpinnings of phylogenetic methods: optimisation algorithms, Monte Carlo sampling, maximum-likelihood estimation, Hidden Markov Model, MCMC algorithm, Bayesian inference, Dynamical programming, traversal algorithms, search algorithms, multiple sequence alignment algorithms.

Extracted sequences from the BLAST database and developed a rigorous phylogenetics pipeline for functional relationships among bacterial transporter molecules for drug target identification.