Chemical Reactivity Landscape, Scientific Project Leader

Position Summary

We are looking for a scientific leader to own the Chemical Reactivity Landscape initiative: a purpose built initiative to build a predictive map of reaction outcomes and generate differentiated reactivity data. This is a hands-on scientific leadership role working at the intersection of quantum chemistry, machine learning, high-throughput experimentation, and laboratory automation. You will set the modeling strategy across mechanistic, quantum chemical, statistical, and learned approaches; direct the synthetic organic core that gives those models something real to learn; and drive the closed loop in which predictions select experiments and experimental results improve predictions. You will also decide how foundation models and agentic systems are used in reaction discovery. The project is run in close partnership with the Synthetic Innovation Team (SIT) in Lilly Small Molecule Discovery (LSMD). The role is on-site in South San Francisco with both people and project leadership scope.

Key Responsibilities

Scientific Direction & Project Leadership

• Own the scientific vision for the Chemical Reactivity Landscape: working with the Synthetic Innovation Team (SIT) in LSMD, define what the landscape must predict, to what accuracy, for which chemistry, and in what sequence it is built
• Translate scientific ambition into a sequenced plan with dated milestones and decision points, and communicate progress, trade-offs, and risk to senior R&D leadership
• Secure and allocate compute, HTE capacity, and headcount against the highest-value scientific questions

Reactivity Science & Modeling

• Partner with computational chemistry, jointly defining the level of theory, conformational sampling, transition-state treatment, and solvation strategy suited to the scientific question at hand
• Drive descriptor and representation strategy so that models generalize across chemotypes and reaction classes, and so that their predictions can be explained in chemical terms
• Lead catalysis-and other reaction discovery and optimization efforts defining the substrate and condition spaces worth searching
• Maintain deep command of the primary literature across organic chemistry, machine learning, and computational chemistry, quickly judging which advances are worth bringing inside

Autonomous Experimentation & Closed-Loop Discovery

• Set the direction for integrating foundation models and agentic systems into reaction discovery, agents that plan calculations, propose conditions, dispatch experiments to automated and cloud laboratories, and interpret the readouts that come back
• Drive the closed loop between prediction and experiment so that HTE campaigns are selected by the model and their results measurably improve it
• Champion low-friction and conversational interfaces that put reactivity models in the hands of bench chemists without requiring them to become modelers
• Partner with automation, HTE, and analytical leadership so that experiments produce structured, model-ready, traceable data by default rather than by retrofit

Evaluation, Rigor & Scientific Standards

• Establish how models and agents are evaluated before they influence experiments: benchmarks, held-out data, fair expert and random baselines, calibrated uncertainty, and documented failure modes
• Treat scientific evaluation of foundation models as a first-class discipline rather than a post-hoc check, and build the harnesses and reference datasets that make it routine
• Establish reproducibility, data provenance, and clear reporting of each model’s domain of applicability

Team Leadership & Development

• Lead, manage, and develop a team of computational and data scientists across career stages, including the Reaction Informatics individual contributors on this project: set clear expectations, give direct feedback, and create the conditions for high performance
• Lead the matrixed cross-functional project team through scientific credibility and clarity of direction rather than authority
• Mentor and coach scientists at all levels, and give people genuine scientific territory to own and grow into
• Recruit, hire, and onboard across career stages, and invest in early-career scientists as part of the long-term talent pipeline for reactivity modeling at Lilly
• Actively build inclusive team relationships and ensure all team members have appropriate access to challenging work, development opportunities, and visibility
• Foster digital fluency beyond the immediate team: encourage experimentation with new tools, support training opportunities, remove barriers to adoption, and recognize scientists who use these capabilities to accelerate real chemistry

Collaboration, Influence & External Engagement

• Partner with the Synthetic Innovation Team, medicinal chemistry, process chemistry, analytical sciences, and automation to turn portfolio bottlenecks into well-posed scientific questions
• Engage LSMD project teams directly to understand their synthetic bottlenecks, and bring an enterprise mindset — proactively connecting teams to reactivity predictions, AI/ML-assisted retrosynthesis, and emerging technologies beyond the immediate scope of any one program
• Represent the project in internal governance bodies, scientific forums, and cross-functional decision-making
• Build and manage external collaborations with academic groups, consortia, and technology partners, and evaluate them for strategic fit rather than novelty
• Publish and present in leading chemistry, cheminformatics, and machine learning venues, and use that visibility to attract talent and partners

Safety, Quality & Compliance

• Keep safety as the top priority; provide guidance on HSE related to new technology deployment, and laboratory operations
• Ensure the team maintains rigorous documentation, data integrity, and compliance standards across all computational and experimental work

What Success Looks Like

• The reactivity landscape is the default reference for condition selection across LSMD, and its uncertainty estimates hold up under real use
• Optimization campaigns reach target conditions in measurably fewer experiments than expert-designed baselines, across multiple reaction classes
• Models generalize to out of domain chemistry, and know precisely where they stop working
• Autonomous and semi-autonomous campaigns run reliably against real instruments, with the loop closing without manual intervention
• Evaluation standards set by this project are adopted by other AI-for-science efforts across Lilly