Applied Scientist II

Coalition, Inc.

5h•Remote

About The Position

We are hiring an Applied Scientist II to build and improve the machine learning and GenAI models that power our underwriting decisions. You will take ownership of high-impact modeling problems end-to-end. This includes framing and data exploration through model design, evaluation, deployment, and monitoring, directly influencing how we assess and price cyber risk. You’ll work closely with underwriters, product managers, and engineers to design robust pipelines, experiment with state-of-the-art ML/GenAI techniques, and ship models that meaningfully move business metrics. Your work will turn complex insurance and security signals into reliable decisioning systems, helping Coalition write better business at scale while pushing the frontier of AI in underwriting.

Requirements

Ph.D. or MS in a quantitative or computational field (e.g., Computer Science, Statistics, Applied Math, Electrical Engineering) or equivalent practical experience.
5+ years of full-time experience developing and deploying ML- and data-based solutions in production.
Practical, hands-on experience with supervised and unsupervised learning methods, including model selection, regularization, and calibration.
Expertise in statistical analysis methods, especially regression analysis, statistical inference, and forecasting / time-series methods.
Strong proficiency in Python and core ML libraries (e.g., scikit-learn, XGBoost/LightGBM, PyTorch/TensorFlow) and SQL for working with large, messy datasets.
Experience with experiment design and evaluation (e.g., A/B tests, offline metrics vs. online KPIs, guardrails) and with ensuring reproducible results.
Comfortable and effective in ambiguous problem spaces; demonstrated ability to own and drive projects with minimal oversight and process.
Exceptional written and oral communication skills, with the ability to explain complex modeling choices and tradeoffs to technical and non-technical stakeholders.
Minimum 1+ year of experience in insurance underwriting modeling (pricing, risk scoring, eligibility, or related applications).

Nice To Haves

Experience with modern GenAI techniques relevant to underwriting (e.g., using LLMs for document understanding, unstructured-to-structured extraction, or underwriter copilot workflows).
Familiarity with model governance in regulated environments, including documentation, validation, monitoring, and change management.
Experience with ML orchestration and MLOps tools (e.g., Airflow, Prefect, MLflow, SageMaker) for managing training, deployment, and monitoring at scale.
Exposure to causal inference or uplift modeling for understanding the impact of underwriting or guideline changes.
Experience working with cyber or P&C insurance data (exposures, limits, deductibles, claims, external risk signals).

Responsibilities

Build and advance our most sensitive and business-critical ML and GenAI models that power underwriting decisions and risk selection.
Drive and execute ML projects end-to-end: problem framing, data exploration, feature engineering, model design, prototyping, offline/online evaluation, deployment, and monitoring.
Design and implement ML pipelines for data preprocessing, feature engineering, model training, hyperparameter tuning, and model evaluation, enabling rapid and reproducible experimentation.
Apply state-of-the-art ML and GenAI workflows (e.g., gradient-boosted trees, deep learning, LLMs, prompt engineering, transfer learning) to improve underwriting accuracy, automation, and decision support.
Own model quality and robustness by defining success metrics, running ablations and diagnostics, and iterating to outperform prior baselines.
Survey and incorporate recent advances in ML/GenAI research into our core underwriting capabilities, balancing scientific rigor with practical constraints.
Collaborate closely with underwriters, product, data, and engineering partners to clarify requirements, align on tradeoffs, and ensure models integrate cleanly into production workflows.
Communicate methods and results clearly through documentation, presentations, and design reviews; share learnings and patterns that level up the broader team.
Contribute to a culture of scientific and data excellence by bringing mature empathy, best practices, and lightweight processes to experimentation, code review, and model governance.