Product-Focused Data Scientist (LLM Products)

Blue Rose Research•New York, NY

6d•$160,000 - $210,000•Hybrid

About The Position

Blue Rose Research is seeking a product-focused data scientist to help build LLM-powered research and analysis tools that deliver greater breadth, depth, and speed. This role involves co-evolving the product experience with the models that power it, aiming to revolutionize how political strategy is developed and communicated to campaigns. The position is for a remote-first company with an office in NYC.

Requirements

5+ years of experience as a data scientist or applied statistician, with lots of exposure to shipping features or products, not just producing analyses.
Proficient in Python, SQL, and standard statistical/ML tooling, and comfortable with the mess of real-world data.
Hands-on experience evaluating or building on top of LLMs: designing eval sets, building LLM-as-judge pipelines, and working with embeddings.
Good product instincts. Cares deeply about how the end user is going to interact with the model through the product.
Communicates clearly in writing and in conversation with engineers, product managers, and non-technical stakeholders alike.
Is energized by ambiguous problems — translating “help win elections” into a model, a metric, or a product decision, without a clean playbook to follow.
Motivated by Blue Rose Research’s mission and interested in learning the progressive political ecosystem.
Authorized to work lawfully in the United States.

Nice To Haves

Prior political or civic-tech experience is welcome but not required.

Responsibilities

Spend time with political strategists to understand how they think and what AI can do to improve their work.
Translate that work into product decisions: how to structure workflows that involve humans and agents working together.
Help set the bar for rigor on what “good” looks like for an AI system, while staying pragmatic about what a real deadline and a real user actually need.
Design evaluations and LLM-as-judge systems that tell us whether an LLM-powered feature is accurate, well-grounded, and useful.
Prototype quickly, using coding agents to test ideas before they become product commitments.
Communicate findings and tradeoffs clearly to both technical and nontechnical audiences, including in client-facing settings.