Actuarial Data Scientist

Shepherd•Chicago, IL

About The Position

Shepherd is building the data infrastructure and predictive models that power modern commercial insurance. As an Actuarial Data Scientist on the Actuarial & Predictive Analytics team, you will own the development of pricing models starting with commercial auto, one of our highest-volume and most data-rich lines. You'll directly shape the quality of the book we write and the products we bring to market. This is a high-impact, individual-contributor role for someone who thrives at the intersection of statistical rigor and shipping real products. You will work closely with actuaries, underwriters, and engineers to turn data into decisions.

Requirements

3+ years of professional experience building and deploying personal auto or commercial lines predictive pricing models in production
Familiarity with actuarial concepts (loss development, exposure rating, credibility)
Strong foundation in statistics: GLMs, GBDTs, time series analysis, heavy tail distributions, and Bayesian methods
Proficiency in Python and SQL
Experience with feature engineering on messy, real-world, small data
Ability to reason from first principles and communicate results crisply to non-technical audiences
AI-native mindset: you already use LLMs and AI tools to accelerate your own work

Nice To Haves

Experience in insurance, insurtech, fintech, or other regulated industries
Exposure to telematics pricing models
Experience with NLP/document extraction from unstructured insurance submissions
Prior work with model deployment infrastructure (AWS)

Responsibilities

Own commercial auto pricing models end-to-end from feature development through deployment and iterate on them as the book grows and new data sources come online
Build and deploy predictive models build and deploy loss cost models that set pricing for Shepherd's commercial auto book
Design and maintain feature pipelines that transform raw submission, claims, and third-party data into model-ready inputs
Collaborate with actuaries and underwriters to translate domain expertise into model features and validate outputs against real-world outcomes
Develop model monitoring frameworks to track drift, performance degradation, and calibration over time
Run experiments and back-tests to quantify model impact on loss ratios, pricing accuracy, and portfolio quality
Communicate findings clearly to technical and non-technical stakeholders through concise documentation and presentations