Kai Armstrong

Independent research and writing on AI for biology, with a focus on evaluation and translation.

• Publish research notes and essays on biological foundation models and AI/biotech industry structure.
• Advise early-stage teams on benchmark design, evaluation strategy, and technical diligence.
• Build open-source tooling for reproducible evaluation of perturbation and single-cell models.

• Evaluation
• Foundation models
• Technical writing

Led evaluation and data work for a platform combining machine learning with high-throughput wet-lab experiments.

• Designed and ran benchmarks for perturbation prediction across internal and public datasets.
• Built data pipelines connecting assay results to model training and evaluation loops.
• Worked with experimental scientists to define endpoints that were both biologically meaningful and measurable.

• Perturbation biology
• Benchmarking
• Data pipelines

Doctoral research on representation learning for single-cell and perturbation data.

• Developed methods for learning structured representations of high-dimensional cellular state.
• Collaborated on benchmark datasets and evaluation protocols used across the lab.
• Published and presented work at machine learning and computational biology venues.

• Single-cell
• Representation learning

Thesis: representation learning for perturbation biology. Minor in computer science.

First-class honours. Undergraduate thesis on probabilistic models of gene expression.