Quantitative Safety Scientist

We are hiring a Quantitative Safety Scientist (Discovery to Post Marketing), who blends statistical methods, real-world evidence, pharmacoepidemiologic judgment, and pharmacovigilance domain fluency. This is deliberately broader than a classical biostatistician: the core challenge in signal detection is not computing a disproportionality statistic but choosing the right method, comparator, stratification, and threshold—and interpreting the result with clinical and regulatory judgment.

You will sit at the intersection of safety science and product, working with engineering to embed defensible methods into the signal and risk modules and with the therapeutic add-on pack roadmap to define class-driven assessments, metrics, and calculations across Graph Clinical Insights.

We are deliberately casting a wider net for this role than a traditional post-marketing signal scientist. The same disciplined thinking that makes a disproportionality method defensible in signal-clear hypotheses, the right comparator, and honest uncertainty is equally valuable earlier in a molecule's life, when target biology, preclinical toxicology, and early clinical data first hint at a safety signal. We are looking for someone who is genuinely energized by that shift-left challenge: comfortable moving between discovery-stage and clinical-stage data as readily as post-marketing case data, and hands-on enough to prototype a new method themselves—in code, not just on a whiteboard—rather than only applying established statistical packages. The goal is a single, reusable body of quantitative methods that travels with a molecule from early identification through clinical development and into lifetime post-marketing surveillance, so insight compounds instead of resetting at each handoff.

Responsibilities:

• Design and validate the disproportionality and statistical signal-detection methods powering /signal - PRR, ROR, EBGM (MGPS), Bayesian Confidence Propagation / IC and IC025, observed-vs-expected, and time-to-onset analyses.
• Own therapeutic-class-driven assessment logic: comparator selection, indication and channeling confounding control, subgroup and class-effect stratification, and the metrics/calculations that differentiate each certified therapeutic add-on pack (Oncology, Anti-Infective/AMR, Cardiometabolic, CNS, Immunology, Respiratory, Vaccines, Rare Disease).
• Shift the aperture left: partner with discovery, translational, and early clinical-development teams to identify candidate safety signals from target biology, mechanism of action, preclinical toxicology, and early-phase clinical data—well before a molecule reaches post-marketing surveillance.
• Design a single, reusable molecule-risk-profile framework that carries structured safety evidence forward across the lifecycle—from discovery and molecule identification, through clinical trial safety analytics, into post-marketing signal and risk management—rather than treating each stage as a separate analytic silo.
• Hands-on prototype of innovative, exploratory e. g., ML-assisted signal triage, biomarker-informed risk stratification, and translational safety scoring—as working proofs of concept in Python/R, ahead of handing validated approaches to engineering for productization.
• Define signal validation, prioritization, and management workflows aligned with GVP Module IX, including SMQ-based screening, signal triage criteria, and signal lifecycle states.

Requirements:

• Requirements are mapped to the product capabilities they enable, so candidates can see exactly why each is needed.
• Master's or PhD in pharmacoeconomics, epidemiology, biostatistics, pharmacy, pharmacology, or a related quantitative life science discipline.
• 4-8+ years in pharmacovigilance/drug safety/RWE (preferred), with hands-on signal detection and management experience.
• Demonstrated expertise in disproportionality analysis: PPR, ROR, EBGM, BCPNN/IC, and their correct interpretation.
• Working knowledge of GVP Module IX (signal management) and ICH E2C/E2E.
• Fluency in MedDRA, SMQs, and WHODrug.
• Proficiency in Python (pandas, numpy, scipy/statsmodels) and/or R for safety analytics.
• Pharmacoepidemiologic judgment; confounding, masking, comparator and threshold selection, false-positive control.
• Clear scientific communication; able to document defensible, auditable methods.
• Innovative, hands-on builder mindset, able to design and prototype novel quantitative methods, not only apply established ones, with genuine curiosity about data from earlier in the molecule lifecycle (discovery, translational, preclinical, or clinical development).

Highly Desirable:

• Risk-management experience—benefit-risk assessment, RMP/REMS, and risk-minimization effectiveness evaluation.
• Experience with FAERS, EudraVigilance, VigiBase, or sponsor safety databases (Argus, ArisGlobal, Veeva Vault Safety).