Machine Learning Lead (LLM) at Blue Rose Research

Job description

About us: Blue Rose Research builds data and AI tools that help Democrats win elections and advance progressive causes. We forecast races, test ads at scale, and use AI to track what's breaking in the news, then help campaigns respond fast with messages that actually move voters. Our work has guided how hundreds of millions of dollars are spent by the country's top campaigns and progressive organizations. We're a small, mission-driven team of engineers, data scientists, and political strategists who build fast, experiment boldly, and care deeply about using technology for good. Machine Learning Lead (LLM) The 2028 cycle starts the day after the midterms, and a wide-open Democratic presidential primary will begin almost immediately. With a crowded field and a fragmented media environment, campaigns will need to understand what's reaching and persuading voters faster than ever, and our tools will be at the center of that work from the first days of the primary through the general election. We're hiring a Machine Learning Lead to guide a small team of senior data scientists building the LLM-powered research and analysis tools behind that effort. The team's core work is making AI systems we can trust and improve. That means training judge models that tell us whether outputs are accurate, well-grounded, and useful, and then using those judges to automatically optimize the context, prompts, and retrieval that power our products. Reporting to the Director of Engineering, you'll own the roadmap, develop the team, and stay hands-on in the technical work. You'll be joining at the start of the cycle, with time to build systems that are ready before the pressure peaks rather than scrambling to catch up. You will: Lead and develop a team of senior data scientists. You'll set direction, run planning and standups, give honest feedback, and help each person grow. Own the team roadmap with senior leadership, prioritizing the work that most improves what strategists and campaigns actually get from our products. Set the bar for what "good" looks like for our AI systems, and build the evaluation infrastructure to measure it: eval sets, human labeling, and LLM-as-judge pipelines. Lead the training and calibration of judge models, making sure they track expert human judgment and stay reliable as products and underlying models change. Build systems that use those judges to automatically optimize context (prompts, retrieval, examples, and agent workflows), so quality improves continuously instead of through hand-tuning. Guide decisions about when to use off-the-shelf models, when to fine-tune open-weight models, and how to get them into production. Keep the team's model work connected to the product by partnering with engineering and political strategists on how real users interact with what we build. Balance rigor with pragmatism, making clear tradeoff calls when a real deadline and a real user are on the line. Communicate results and tradeoffs clearly to technical and nontechnical audiences, including in client-facing settings. Our ideal candidate likely: Has 6+ years of experience in ML, data science, or applied statistics, including 1+ years leading a team as a manager or tech lead. Has shipped ML or LLM-powered features to production, not just research prototypes or analyses. Has a strong grounding in applied statistics. Is proficient in Python, SQL, and modern ML frameworks (PyTorch, JAX, or similar), and comfortable with cloud and production tooling. Has good product instincts and cares about how the end user will experience the model through the product. Is a leader people want to work for: clear, direct, invested in teammates' growth, and able to keep a team focused when priorities shift quickly. Is energized by ambiguous problems, like translating "help win elections" into production models. Join us! We are a remote-first company with an office in NYC and regular team meetups. We primarily work East Coast hours. The salary range for this position is $165,000 - $210,000 annually, commensurate with experience. We offer medical, dental, vision, 401k match, 20 weeks paid parental leave, fertility and family-building support through Maven Clinic, and unlimited PTO. You don't need to meet every qualification to apply. We welcome applicants from all backgrounds. Candidates must be authorized to work lawfully in the United States.

Requirements

Must-haves

Lead and develop a team of senior data scientists. You'll set direction, run planning and standups, give honest feedback, and help each person grow.
Own the team roadmap with senior leadership, prioritizing the work that most improves what strategists and campaigns actually get from our products.
Set the bar for what "good" looks like for our AI systems, and build the evaluation infrastructure to measure it: eval sets, human labeling, and LLM-as-judge pipelines.
Lead the training and calibration of judge models, making sure they track expert human judgment and stay reliable as products and underlying models change.
Build systems that use those judges to automatically optimize context (prompts, retrieval, examples, and agent workflows), so quality improves continuously instead of through hand-tuning.
Guide decisions about when to use off-the-shelf models, when to fine-tune open-weight models, and how to get them into production.
Keep the team's model work connected to the product by partnering with engineering and political strategists on how real users interact with what we build.
Balance rigor with pragmatism, making clear tradeoff calls when a real deadline and a real user are on the line.
Communicate results and tradeoffs clearly to technical and nontechnical audiences, including in client-facing settings.
Has 6+ years of experience in ML, data science, or applied statistics, including 1+ years leading a team as a manager or tech lead.
Has shipped ML or LLM-powered features to production, not just research prototypes or analyses.
Has a strong grounding in applied statistics.
Is proficient in Python, SQL, and modern ML frameworks (PyTorch, JAX, or similar), and comfortable with cloud and production tooling.
Has good product instincts and cares about how the end user will experience the model through the product.
Is a leader people want to work for: clear, direct, invested in teammates' growth, and able to keep a team focused when priorities shift quickly.
Is energized by ambiguous problems, like translating "help win elections" into production models.

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aimachine learningllmmlpythonsql

Platform: greenhouse
Effort: quick

Benefits

Benefits detected in the listing.

Dental

Vision

401k

Pto

Parental Leave