Principal Scientist, Data Science

Johnson & Johnson Johnson & Johnson · Pharma · Titusville, NJ +5

Principal Scientist role focused on leveraging Real-World Data (RWD) for clinical trial design, feasibility, and execution. This involves applying ML, optimization, and GenAI techniques, including LLMs for data extraction and structured analysis, to inform protocol decisions, operational plans, and patient journey simulations. The role requires strong programming skills, MLOps experience, and familiarity with healthcare data and privacy.

What you'd actually do

  1. Use RWD to quantify disease prevalence, care pathways, and the impact of inclusion/exclusion criteria; produce feasibility scoring across geographies, sites, and subpopulations.
  2. Assess RWD‑feasible endpoints and proxies; evaluate availability, completeness, quality, and signal‑to‑noise to guide protocol design choices.
  3. Construct RWD‑based cohorts and external/synthetic controls to benchmark protocol decisions and stress‑test sample‑size/timeline assumptions.
  4. Develop ML and multi‑objective optimization solutions primarily powered by RWD to surface trade‑offs (speed, quality, cost, diversity) and recommend design and operational scenarios informed by real‑world care patterns.
  5. Build RWD‑calibrated stochastic simulations of patient journeys to forecast timeline sensitivities and completion risk; provide RWD features, calibration sets, and feasibility constraints to the partner team’s enrollment/screen‑failure/retention models.
  6. Adapt LLMs/GenAI for structured extraction from RWD artifacts (structured and unstructured EHR, notes, radiology/pathology reports, claims, registries); harmonize concepts to standard vocabularies to support eligibility criteria evaluation and schedule‑of‑activities insights grounded in real‑world practice.
  7. Clearly communicate RWD‑based assumptions, methods, and results to clinical, operational, and leadership stakeholders; coach and mentor colleagues on RWD methodologies, pipelines, and best practices.

Skills

Required

  • Ph.D. degree in quantitative discipline (e.g., computer science, electrical and computer engineering, biostatistics, health economics, biomedical informatics, applied mathematics, or similar)
  • 5+ years delivering ML/NLP/GenAI and multi objective optimization solutions with primary reliance on RWD (EHR, claims, registries, digital health), including collaboration with operations analytics teams.
  • Hands on experience with multimodal RWD (structured + unstructured) predictive modeling and stochastic simulations for feasibility and time series scenario forecasting.
  • Demonstrated ability to construct, validate, and deploy models from RWD to inform trial feasibility, endpoint selection, eligibility criteria effects, and external control design.
  • Experience building optimization engines (e.g., evolutionary algorithms, reinforcement learning, mixed integer linear programming) using RWD derived signals to navigate complex tradeoffs.
  • Proficiency in MLOps (e.g., MLflow, Kedro), Git, and CI/CD; strong programming skills in Python and SQL; familiarity with DSPy/LangChain, pymoo, scikit learn, XGBoost, Optuna, PyMC.
  • Familiarity with healthcare privacy/compliance, de identification practices, and RWD data quality management; ability to integrate outputs from operational systems/models when needed while keeping the analytical core RWD driven.

Nice to have

  • Demonstrated expertise applying RWD methods to inform trial design: target trial emulation, propensity weighting/matching, and survival/time‑to‑event analyses for endpoint feasibility and external/synthetic controls.
  • Proven collaboration with operations analytics teams by supplying RWD‑derived cohorts, features, and feasibility evidence that improved enrollment forecasting, site selection, and diversity goals.

What the JD emphasized

  • primary reliance on RWD
  • ML/NLP/GenAI and multi objective optimization solutions
  • multimodal RWD (structured + unstructured) predictive modeling
  • construct, validate, and deploy models from RWD
  • building optimization engines
  • MLOps
  • Python and SQL
  • DSPy/LangChain

Other signals

  • Lead analytics, ML, optimization, and GenAI that rely primarily on real-world data to inform clinical trial design, feasibility, and execution facilitation
  • Adapt LLMs/GenAI for structured extraction from RWD artifacts
  • Develop ML and multi-objective optimization solutions primarily powered by RWD