Deep Origin is a biotechnology company building an operating system for science that transforms how life science research is conducted. Our AI-powered platforms simulate biology, streamline R&D workflows, and enable scientists to model complex biological systems at scale.

Deep Origin leads the PREDICTS consortium, funded by ARPA-H under the CATALYST program, to build in silico models that predict human drug safety and toxicity. As part of this work we are scaling our capabilities in whole-human modeling and predictive toxicology.

About the role

Deep Origin is building VHAT, the Virtual Human Avatars of Toxicology. VHAT takes a molecule, a dose and a patient population and predicts organ-level human toxicity before a compound is synthesized or dosed. Metabolism sits at the heart of that chain. What a drug becomes, which enzymes act on it and how fast they act determine what each organ is exposed to, and every predicted metabolite is screened for its own off-target and organ liabilities. The Metabolism group builds four connected models: drug-metabolizing enzyme, site of metabolism, metabolite structure, and enzyme kinetics. As the group’s leader, you will own those models, mentor the team that builds them, and drive integration of their predicted parameters into the physiologically based pharmacokinetic (PBPK) model that governs exposure across the VHAT.

Applicants must be authorized to work for any employer in the U.S. We are unable to sponsor or take over sponsorship of an employment Visa at this time.

Requirements

• A PhD in computational chemistry, cheminformatics, drug metabolism, computational biology, pharmacology or a related field.
• Deep, hands-on expertise in drug metabolism. You understand CYP and non-CYP enzymes, sites and products of metabolism, enzyme kinetics and clearance, and you can judge whether a prediction is chemically and biologically sound.
• A strong track record in computational modeling, either as a model builder or as an expert end-user. This includes machine learning for molecular property or reactivity prediction, and mechanistic or kinetic modeling, including ODE-based models, parameter estimation, and Bayesian or frequentist inference. You work fluently in Python.
• Demonstrated rigor in model validation. You have designed appropriate data splits, defined applicability domains, and run external or blinded benchmarks, and you can explain the limits of a model as clearly as its strengths.
• Experience leading a scientific team and delivering against milestones and deadlines.
• Clear written and spoken communication with scientists, partners and funders.

Strongly preferred

This role owns the link between metabolism and whole-body exposure, so PBPK experience will weigh heavily in our decision.

• Hands-on experience building or parameterizing PBPK models in Simcyp, GastroPlus, PK-Sim, SimBiology or a custom ODE framework.
• Experience translating in vitro or predicted intrinsic clearance into in vivo hepatic clearance and human PK, including the use of IVIVE scaling factors and an understanding of where they tend to fail.
• Experience using PBPK to assess metabolic drug-drug interactions, such as CYP inhibition and induction.
• Experience combining machine learning predictions with mechanistic PK models, and carrying their uncertainty into downstream predictions.
  

Nice to have

• Experience with metabolomics or LC-MS metabolite identification data.
• Experience running data curation at scale, including annotation standards and quality control.
• Experience on government-funded research programs with milestone-based reporting.
• Exposure to regulatory science, such as FDA guidance on new approach methodologies, model-informed drug development or the ISTAND program.
• Familiarity with medicinal chemistry and how metabolism liabilities are addressed during lead optimization.

Responsibilities

• Lead a team of computational scientists and data curators across sites. Set scientific direction, priorities and review standards for each model, and develop the people who build them.
• Own the roadmap and release gates for the four metabolism models. Decide when a model is ready for the VHAT platform, for use as a standalone product, and for external review.
• Own the metabolism-to-PBPK integration. Define which parameters pass to the PBPK model, how predicted or in vitro clearance is scaled to in vivo hepatic clearance, and how the combined model is checked against clinical PK and drug-drug interaction data. Work with PBPK modelers so that parent and metabolite exposure flow correctly into the organ models.
• Design validation experiments that hold up to outside scrutiny. This includes leakage-aware data splits, applicability-domain and confidence reporting, predictions locked before new experimental data arrive, and external, blinded and independent benchmarks.
• Set the data strategy for a field where high-quality labels are scarce. Plan curation targets and throughput, oversee AI-assisted literature mining with expert quality control, support data licensing, and design wet-lab experiments with partners (recombinant enzymes, human liver microsomes, hepatocytes), including active-learning compound selection.
• Extend coverage beyond CYP Phase I metabolism to Phase II enzymes (UGT, SULT), FMO, AOX and multistep metabolic networks, and scope how enzyme polymorphism and population variability enter the models.
• Manage external relationships with consortium partners, contract research organizations, academic reviewers and independent validation teams.
• Author technical milestone reports and present results to ARPA-H. Contribute metabolism content to grant proposals and to regulatory engagement on new approach methodologies.
• Work with the modeling and integration teams to carry uncertainty from the metabolism models through PBPK to the final organ-level risk.

What success looks like

In the first 6 months, you will have taken ownership of the four models and the team. Metabolite identity and enzyme-isoform predictions will be released on the VHAT platform and reported to ARPA-H. Predicted clearance and kinetic parameters will be passed into the PBPK model with documented methods and scaling assumptions. Prospective predictions on new wet-lab metabolism data will have been locked in advance and scored. A first version of the kinetics model will be in place, and you will have set a clear plan for curation throughput, enzyme coverage and population variability.

By 12 months, the metabolism models will have been tested by an independent validation team on held-out data. Metabolite concentration profiles will feed PBPK for both parent drugs and metabolites, and from there the organ models. Coverage will extend beyond CYP Phase I enzymes, and confidence estimates from the metabolism models will carry through to VHAT risk outputs. The team will be running on a roadmap you set, with each model's strengths and limits clearly documented.

Why join us

This is a chance to lead metabolism science inside a program that links atomic-level predictions to whole-human safety outcomes. The work is hard because the data are scarce, and it matters because toxicity remains a leading cause of clinical failure.

Benefits

• Competitive compensation package with meaningful equity.
• Comprehensive health, dental, and vision coverage.
• A remote-friendly culture across the U.S. and Europe, with flexible, cross-time-zone working.
• Opportunity to shape the future of health, longevity, and our ability to simulate life.
• Annual team gatherings and company events.