Machine Learning Scientist – Clinical Prediction

Iambic Iambic · Pharma · Boston, MA · Technology

Machine Learning Scientist to fine-tune a multimodal transformer model for clinical and biomedical applications in drug discovery. The role involves data curation, model fine-tuning, experimental design, evaluation, and collaboration for production deployment.

What you'd actually do

  1. Fine-tune large-scale multimodal transformer models for clinical and biomedical applications
  2. Identify, characterize, and utilize datasets that can deliver insights into pharmacokinetics (PK), pharmacodynamics (PD), toxicity, clinical adverse events, and clinical trial outcomes
  3. Develop and apply rigorous experimental approaches that account for multiple sources of potential leakage (split, metadata, trial-family, temporal, ontological, arm-comparator, etc.)
  4. Design and maintain benchmarking and evaluation frameworks that track model quality across models and tasks
  5. Build models with appropriate calibration, uncertainty quantification, and clinically meaningful evaluation metrics.

Skills

Required

  • MS in chemistry, bio/chemical engineering, or a computational STEM field with 3+ years of relevant industry or research experience, or PhD or equivalent industry experience demonstrating comparable depth
  • Strong Python experience, including implementing and fine-tuning deep learning models
  • Demonstrated experience in clinical science or working with clinical datasets
  • Excellent Data Science skills (problem framing, data sourcing, extraction, cleaning, visualization, EDA, modeling, tuning, storytelling, etc.)
  • Enough independence to own a workstream from data ingestion through evaluation
  • Strong engineering habits: reproducible experimentation, appropriate control strategy, clean code, testing
  • Comfort working with modern ML infrastructure (e.g., Docker, CUDA, Kubernetes, experiment tracking such as Weights & Biases)

Nice to have

  • Experience building and deploying clinically relevant prediction models
  • Familiarity with [ClinicalTrials.gov/AACT](http://ClinicalTrials.gov/AACT) data
  • Experience with MedDRA, pharmacovigilance, or adverse event data
  • Direct exposure to multi-task learning
  • Hands-on experience with agentic data extraction
  • HPC or large-scale computing experience

What the JD emphasized

  • clinical fine-tuning
  • multimodal transformer model
  • rigorous, leakage-resistant experimental frameworks
  • clinically relevant prediction models

Other signals

  • fine-tuning multimodal transformer models
  • clinical prediction
  • drug discovery
  • production deployment