PhD Research Internship

AI for drug design is starting to have real world impact but most companies struggle deploying AI effectively as they lack data and expertise. deepmirror is a drug design Foundation Model built from curated, non-public experimental molecule property measurements (potency, binding, ADMET) aggregated from patents, papers, and partners. We help chemistry teams choose the next best experiments to progress programs faster with fewer dead ends. Since launching in 2023, our platform is now used by hundreds of chemists across the globe to impact real world drug programs in oncology, dementia, inflammation, and global health.

In this internship, your primary focus will be on intelligent compound selection for ADMET modelling specifically, how to choose the smallest, most informative set of compounds to measure in order to build predictive models that generalise well across a defined chemical space. The role combines rigorous benchmarking with practical impact, as you develop and evaluate strategies that maximise model performance under realistic experimental budget constraints.

As part of our product-focused team, you will bridge the gap between academic research and industrial applications, translating insights from active learning and cheminformatics into practical guidance for drug development workflows. We maintain a strong commitment to scientific excellence and actively support publication of research findings in peer-reviewed journals.

You will:

• Design and benchmark compound acquisition strategies, including diversity-based, uncertainty-based, and hybrid active learning approaches
• Evaluate model generalisation across ADMET endpoints in realistic low-data settings
• Investigate how data quality interventions (e.g. removal of activity cliffs, outliers, or noisy datapoints) affect predictive performance
• Publish results in a scientific journal

Strong understanding of cheminformatics and machine learning, with practical experience applying these to real-world problems. Experience in Python, RDKit, and PyTorch. At the time of internship, you must be enrolled as a Ph.D. student at a University. Willingness to work in-person in London for the duration of the internship.

Nice to Have:

• Experience curating and preparing datasets for computational chemistry or ADMET modelling
• Familiarity with active learning methods or Bayesian optimisation
• Familiarity with QSAR/ADMET modelling and molecular property prediction
• Experience working with public chemical databases (ChEMBL, PubChem)

If you meet at least 60% of the requirements or nice-to-have qualifications, we encourage you to apply.

Competitive salary - paid internship. Social events. Central London Offices.