Data Scientist- Underwriting Analytics

Novacore•Conshohocken, PA

28d•Hybrid

About The Position

Novacore's actuarial team owns the number: they set each program's loss cost, rate need, and target loss ratio. This role owns what happens next. You'll build the model that distributes the actuarial loss cost across real risk characteristics (building class, occupancy, TIV, geography, and beyond) and then design the underwriting levers that let the business hit that target loss ratio while staying competitive: guideline gates, segmentation cuts, terms, and recommendations to actuarial on new risk factors to consider for pricing. You're not setting program profitability. Actuarial does that. You're finding creative, defensible ways to make the program more profitable within it. This isn't a pricing-actuary role, and it isn't a reporting/BI role either. It's a hybrid that sits between actuarial output and underwriting action: part pricing analytics, part underwriting partnership, built to own a cluster of programs rather than one at a time. This is a hybrid opportunity from the Conshohocken, PA Home Office

Requirements

Strong grounding in mathematical statistics, actuarial science, and practical insurance/underwriting fundamentals.
Data science, Actuarial or actuarial-adjacent background (2+ actuarial exams passed, or an equivalent quantitative pricing credential). You'll work constantly with actuarial output, but this is not an actuarial pricing seat.
P&C insurance experience, commercial lines and/or MGA/program business preferred.
Strong SQL and Python, with hands-on GLM / predictive modeling experience applied to underwriting or pricing, not exclusively reserving.
Familiarity with modern data tools: version control (Git), cloud platforms, BI tools (Power BI, Tableau), or notebook environments (Jupyter).
You already use AI tools in your daily workflow.
Curiosity about how LLMs and agentic frameworks can automate guideline codification and gap-analysis work.
Willingness to prototype, experiment, and ship, not just theorize.
A track record of presenting analytical findings to underwriters and getting them adopted, including handling pushback, across more than one stakeholder relationship at a time.
Comfort operating in ambiguity.

Nice To Haves

You've built a GLM in Python or R and watched an underwriter actually change a decision because of it, not just admired the fit statistics.
You've turned a rate need into an actual guideline or appetite change, not just a slide.
You've used an LLM to speed up something tedious (guideline extraction, report drafting) and it changed how you think about your job.
You want to build the underwriting decision layer for a cluster of programs, not just run someone else's rating plan.

Responsibilities

Translate underwriting guidelines for every program, currently spread across PDFs, institutional memory, and individual underwriters' heads, into structured, version-controlled rule sets.
Build gap analyses comparing what guidelines say against what's actually being bound, priced, and retained.
Take the actuarial team's loss cost and target loss ratio for a program and partner with actuarial to build underwriting models that allow program underwriters to find the best quality risks.
Apply modern data science techniques (predictive modeling, GLMs, gradient boosting, clustering) to build underwriting models and find segmentation opportunities the current rating plan misses.
Leverage AI tools (LLMs, agentic workflows, Claude) to automate and augment guideline-codification and gap-analysis work.
Translate the underwriting models into clear underwriting action: eligibility gates, class-level appetite, terms (limits, exclusions, deductibles, endorsements), and pricing flexibility (within approved actuarial guidelines) that hit the target loss ratio while staying competitive.
Partner directly with your underwriting leaders to socialize findings and get guideline or lever changes adopted, not just documented.
Feed adopted changes and their measured impact back into the actuarial team's next loss pick. This is a two-way loop, not a one-way handoff: you don't override the actuarial number; you make sure underwriting can act on it.
Document findings and adopted changes so institutional knowledge compounds instead of living in one person's head.
Build reusable frameworks and tooling that extend across every program, not one-off analyses.
Partner with the product team and actuarial team to form a cross functional pod that collectively owns the profitability of assigned programs. You will exist as one team with one voice. Collaboration, partnership and team work are critical to the success of the pod.