Advisor - Reaction Informatics, Chemical Reactivity Landscape

About the Lilly and NVIDIA Partnership

Lilly and NVIDIA are launching a new AI co-innovation lab in the heart of Silicon Valley — an up-to-$1 billion, multi-year commitment to solve drug discovery’s toughest challenges. The lab brings Lilly scientists, technologists, chemists and biologists together with NVIDIA engineers under one roof. Together, we are building purpose‑built foundation and frontier AI models trained on Lilly data at scale, tightening the feedback loop between automated wet labs and computational dry labs, designing the next generation of medicines for millions of patients across the globe.

The Chemical Reactivity Landscape project is building a quantitative, predictive map of how reaction outcome depends on substrate, catalyst, and conditions across the chemistry Lilly runs. We are looking for a scientist to own the modeling and analysis layer of that map: turning high-throughput experimentation (HTE) and reaction condition data into models that tell project chemists which conditions to run next, and why.

This is a hands‑on individual contributor role reporting to the Scientific Project Leader for the Chemical Reactivity Landscape project. You will work at the interface of quantum chemistry, machine learning, and experimental reaction data, computing descriptors that encode steric and electronic effects, fitting and validating models against real plate data, and closing the loop between what can be calculated about a molecule and what is observed in the lab.

You will be collocated with chemists who run the automation platform to quickly assess the quality of your models and how the analytical readouts behind them are produced. The role is primarily computational and lab‑adjacent, based on‑site in South San Francisco. You will own significant components of the reactivity modeling stack and be trusted to make technical decisions within them.

• Reaction Data Modeling & Analysis Build and validate models that predict reaction outcome, yield, selectivity, conversion, impurity profile from substrate structure, catalyst and ligand identity, and condition variables. Analyze HTE datasets end to end: representation choice, feature engineering, cross‑validation design, uncertainty quantification, and honest out‑of‑domain assessment. Turn plate‑level output into structured, model‑ready reaction records with consistent condition encodings, so that campaigns compound into a reusable reactivity dataset rather than isolated screens.
  
  Characterize what the models do not know: identify under‑sampled regions of condition space and specify the experiments that would most reduce uncertainty. Detect and diagnose reactivity cliffs and mechanistic regime changes in the data rather than smoothing them away.
• Quantum Mechanical & Physical Organic Modeling Run DFT and semi‑empirical calculations on substrates, catalysts, and intermediates to quantify steric and electronic effects and to assess whether a transformation is energetically favorable. Generate atom‑, bond‑, and molecule‑level QM descriptors and combine them with structural representations so that models are grounded in physical organic chemistry rather than correlation alone. Handle conformational sampling, barrier and transition‑state estimation, and solvation and counter‑ion effects at the level of theory the question actually warrants.
  
  Bridge from what is calculated about a molecule to what is measured in the plate, and treat the mismatch between the two as information about the model, the mechanism, or the measurement. Automate QM workflows so descriptor generation scales to thousands of structures without hand‑holding.
• Analytical Chemistry & Data Quality Work fluently with the analytical readouts…