Charter
Be a founding member of the team building the first accurate AI systems for replacing animal and legacy toxicity experiments with human-relevant predictive models.
You will help answer one of the hardest questions in drug discovery:
Given a molecule’s structure, potency, exposure, and biological response, will it be toxic in humans — and why?
About Axiom
Axiom is building a compounding ecosystem to replace animal testing and, over time, reshape how clinical trials are run. It starts with deeply understanding the needs of drug hunters inside large pharma. Those needs shape the world-class datasets we build from scratch. We then use that data to advance our own ML research, while also collaborating with leading AI labs to improve frontier models’ ability to reason over Axiom’s data inside Axiom’s agent harness. This creates a compounding loop: deeper customer understanding shapes the data we generate; better data improves frontier models, Axiom’s fine-tuned models, and our agentic infrastructure; stronger models and tooling expand the capabilities we can offer; and those capabilities are forward deployed into pharma's drug discovery workflows, where scientists use them to solve the highest value drug discovery problems. In turn, this helps us identify the next problems to tackle. Today, we are focused on solving drug-induced liver injury through an integrated data and agentic system already being used by 7 of the top 20 pharma companies and several of the world’s most innovative biotechs. Over time, Axiom will build the world’s largest human datasets across all the major organ systems, paired with an agentic harness that uses this data to predict human drug outcomes dramatically better than animals.
What you will do
You will help define Axiom’s core ML research agenda and build the models that power our product.
Define end-to-end ML and agent systems spanning wet-lab data generation, data cleaning, feature extraction, representation learning, model training, evaluation, inference, deployment, and customer-facing outputs.
Build novel models that learn the relationship between chemistry, biological response, dose, exposure, and human toxicity.
Train large multimodal models on paired chemical structures, high-content cellular images, transcriptomics, proteomics, mass spectrometry, ADME, and clinical outcome data.
Develop foundation models and representation-learning systems for biological images, molecules, and multimodal experimental readouts.
Architect models that predict human toxicity as a function of dose, Cmax, in vitro potency, chemical structure, and biological state.
Develop new ways to aggregate, pool, align, and interpret embeddings across assays, doses, timepoints, modalities, compounds, and biological systems.
Work on contrastive learning, self-supervised learning, semi-supervised learning, multimodal learning, graph neural networks, biological image models, generative models, and mechanistic reasoning systems.
Build models that can generalize across chemical space, mechanisms, targets, assays, and customer programs.
Conduct rigorous error analysis to understand when models fail, why they fail, and what data would make them better.
Collaborate with computational biologists, chemists, mass spec scientists, data engineers, and wet-lab teams to design experiments that maximally improve model performance.
Help build Axiom’s mechanistic agents: systems that reason over experimental data, compare compounds to mechanistic neighbors, explain toxicity mechanisms, and guide scientific decisions.
Own the research-to-product loop: prototype, train, evaluate, ship, observe real usage, improve, and repeat.
Ship insanely great models and products to customers.
We are especially interested in people excited by:
Multimodal ML across chemistry, cellular imaging, transcriptomics, proteomics, mass spectrometry, ADME, and clinical outcomes.
Reasoning over massive amounts of multimodal experimental data, model outputs, literature, and mechanistic evidence.
Reinforcement learning for basic biology, chemistry, and advanced drug discovery.
Self-supervised and semi-supervised learning on high-content imaging and biological readouts.
Uncertainty estimation, calibration, and confidence for scientific decision-making.
Mechanistic interpretability for biological and chemical models.
Evaluation systems for models that must perform on real drug discovery problems, not toy benchmarks.
What we are looking for
We are looking for someone with exceptional ML talent, strong engineering ability, and the ambition to become a leader in AI for biology and drug discovery.
You might be a great fit if:
You have done at least one piece of work, in industry, academia, open source, or independently, that shows exceptional machine learning ability.
You are deeply technical and comfortable writing PyTorch, debugging training runs, working with messy data, scaling inference, and building real systems.
You are excited by non-standard, thorny modeling problems where the data is noisy, multimodal, sparse, biased, biological, and deeply important.
You want to work on ML problems where better models can directly change scientific and clinical decisions.
You are not afraid of the data dirty work required to make models better.
You can move between research ideas and production systems.
You care about evaluation, calibration, failure modes, and real-world usefulness.
You are curious enough to learn biology, chemistry, toxicology, pharmacology, and drug discovery.
You want to grow as both a researcher and an entrepreneur.
You want your work to become a product that customers love and rely on.
Technical skills we value
We do not expect every candidate to have all of these, but we are especially excited by experience with:
PyTorch, JAX, TensorFlow, or other deep learning frameworks.
Python, NumPy, Pandas, Polars, PyArrow, scikit-learn, and scientific computing.
Training and evaluating deep neural networks at scale.
Representation learning, embeddings, contrastive learning, metric learning, and self-supervised learning.
Computer vision models, especially for biological imaging, microscopy, cell painting, or high-content screening.
Multimodal ML across images, molecules, text, omics, mass spec, or tabular data.
Large-scale model training, distributed training, GPU infrastructure, inference pipelines, and cloud compute.
Model evaluation, ablations, benchmarking, uncertainty estimation, calibration, and interpretability.
LLMs, agents, retrieval, tool use, and reasoning systems.
Biology, chemistry, toxicology, pharmacology, or drug discovery datasets.
