Computational Scientist (Biology)

About Axiom:

Axiom’s mission is to end unexpected toxicity and make it possible to test drugs in humans rapidly, at massive scale, and without exposing patients to unforeseen risks. We are starting with liver toxicity, one of the leading causes of drug failure. We’ve built an extraordinarily accurate, high-throughput human liver safety platform that combines advanced human cell systems, imaging, transcriptomics, mass spectrometry, and drug metabolism measurements. Our AI agents integrate this experimental data with chemical structure, human exposure, and thousands of clinical reference compounds to predict whether a drug will be toxic in humans, why, and how to redesign it to be safer.

We work with seven of the world’s top twenty pharmaceutical companies to bring these capabilities directly into drug discovery, enabling scientists to design safety into molecules rather than discover toxicities years later. We also partner with leading frontier AI labs to turn our drug discovery expertise and experimental systems into training environments that teach AI models to become superhuman drug hunters. We then deploy these frontier models into applications that extend across all of drug development: making cancer therapies dramatically safer, enabling treatments for chronic diseases that demand exceptional safety, and rescuing promising drugs previously abandoned because of toxicity. Our ultimate ambition is to build a superhuman AI agent that can predict what will happen to a drug in humans before it ever enters a patient.

What you will do:

• Drive the exploration and analysis of a large multimodal toxicity dataset, turning noisy, high-dimensional imaging, transcriptomic, proteomic and biochemical data into clear biological insights and model-ready features.

• Analyze high-content imaging and transcriptomic data from primary human hepatocytes and multicellular liver systems, identifying phenotypes such as mitochondrial dysfunction, cholestasis, lipid accumulation and ER stress.

• Develop rigorous biological QC methods to assess reproducibility, batch effects, assay performance, and experimental variability across multimodal datasets.

• Collaborate with ML scientists to improve models and agentic workflows that predict drug toxicity and mechanisms of injury, including analyzing what the models capture and where they fail.

• Investigate model behavior, interpret learned biological signals, and systematically identify failure modes.Identify new modalities that could fill the gaps, and lead pilots to test them.

• Work closely with wet lab scientists on experimental protocols and new assays optimized for predictive modeling, not just biological plausibility.

• Partner with leading pharma and biotech teams to interpret molecule toxicity profiles and help them understand the biology driving model predictions.

• Help invent the future of computational toxicology: AI systems that do not just classify compounds, but explain mechanisms, reason over evidence, and guide better drug design.

What we look for:

We are looking for someone who is unusually strong at both biology and computation.

• You are a biologist with exceptional computational skills, or a computational scientist who loves working close to raw experimental data rather than abstracted datasets.

• You are exceptional at finding signal in messy biological data, with strong statistical intuition and the ability to distinguish real biological insights from noise, confounding, artifacts, and overfit correlations.

• You are obsessive about data quality, reproducibility and scientific rigor.

• You can bridge wet lab protocols and machine learning, and you understand how experimental design, assay biology, feature extraction and modeling choices affect one another.

• You are excited by high-content imaging, transcriptomics and assay development, and by working across many toxicity mechanisms, from mitochondrial toxicity and cholestasis to reactive metabolites, transporter biology, and immune, kidney and cardiac toxicity.

• You are excited to evaluate how AI models and agents reason over biological evidence, identify their failure modes, and shape the experimental data needed to improve them.

Technical skills we value:

We do not expect every candidate to have all of these, but we are especially excited by experience with:

• Strong Python programming, Pandas, NumPy, SciPy, scikit-learn, reproducible analytical workflows, and scientific software development

• Statistical inference, experimental design, variance decomposition, dose-response modeling, dimensionality reduction, regression, classification, and rigorous model evaluation

• High-content imaging analysis, microscopy, morphology profiling, and image-based phenotyping

• CellProfiler, Cellpose, napari, OpenCV, scikit-image, or related image analysis tools

• Transcriptomics, proteomics, mass spectrometry, ADME, or other high-dimensional biological datasets

• High-throughput screening, assay development, automation, and experimental QC

• Biological interpretation of model outputs and error modes

• Scientific storytelling: turning complex analyses into clear, credible, inspiring narratives