> whoami

Jonathan Harrison

Senior AI Research Scientist @ Recursion

Machine learning for drug discovery. Building models and tools that make finding new medicines more efficient. Oxford maths & systems biology background.

GitHub CV

cat about.md

I'm a senior AI research scientist at Recursion, where I build machine learning models for drug discovery. My work spans molecular property prediction, uncertainty quantification, and active learning — building tools that help drug designers find new medicines more efficiently.


Before industry, I studied mathematics at Oxford (MMath, 1st class) and completed a DPhil in Systems Biology, applying Bayesian inference and stochastic modelling to problems in cell biology. I bring a rigorous quantitative foundation to applied ML problems in pharma.

cat experience.yml

2024 — Present
Senior AI Research Scientist
Recursion
  • Training and fine-tuning foundation models for molecular property prediction
  • Analysis cited in Recursion 10-K: ">2.5x increased efficiency in detecting new bioactive scaffolds with >40% reduction in flagging of likely cytotoxic compounds"
  • Uncertainty quantification for ML models with applications in Bayesian Optimization and Active Learning
2022 — 2024
AI Research Scientist
Exscientia
  • Led product team for molecular property prediction
  • Core contributor to Molflux, an open-source ML ecosystem for chemistry
  • ML models and active learning software used day-to-day by dozens of drug designers
2019 — 2022
Postdoctoral Research Fellow
University of Warwick
  • Modelled chromosome dynamics using Bayesian statistics and stochastic differential equations
  • Developed computational tools providing insight into how and why cell division goes wrong
  • Supervised MSc rotation projects for 2 PhD students
2018
Data Science Intern
Hudl
  • Event detection in football matches using deep learning

ls ~/projects/

Molflux

Core contributor to an open-source ML ecosystem for chemistry. Provides tools for molecular featurisation, model training, and deployment. Used in production by drug design teams.

open-source python ml chemistry

Polaris Competition

Achieved 2nd place in a public molecular property prediction benchmark competition (2025). Demonstrated state-of-the-art predictive modelling on real-world drug discovery datasets.

competition molecular-ml 2nd-place

cat skills.json

// languages
"expert": ["Python", "Julia"],
"proficient": ["C++", "R", "Stan"],
"tools": ["Git", "Linux", "Docker"],

// domains
"ml": ["Molecular Property Prediction", "Foundation Models", "Uncertainty Quantification"],
"drug_discovery": ["Active Learning", "Bayesian Optimization", "Cheminformatics"],
"statistics": ["Bayesian Inference", "Stochastic Modelling", "MCMC"]

git log --oneline

→ Full list on Google Scholar

echo $CONTACT