Data Scientist | Modeling

About the Role

Our team builds machine learning models for some of the largest health plans in the country to identify and audit claims errors from simple errors to outright fraud. We are looking for a Data Scientist to advance our models further. In addition to building best-in-class models, this person will create technical frameworks and tools to help the team achieve greater scale and improved business outcomes.

What You’ll Do

• Master Claim Audits: Dive into various clinical and coding audits, unpacking the data that drives audit decisions, and understanding past outcomes.
• Model Advancement: Curate labeled data, hypothesize new features, and develop ML models that target claims for audit with even greater precision.
• Optimize Impact: Measure model performance rigorously, translating outcomes into operational insights and recommendations for our clients.
• Tame Complexity: Work with vast, nuanced data pipelines derived from intricate business workflows—often unclean and undefined—to refine them for robust modeling.
• Drive Team Success: Enhance data pipelines, modeling infrastructure, and team tools to level up our capabilities and decision-making efficiency.
• Grow with Healthcare: Build expertise in healthcare data and make an impact on the industry. Engage with clients and internal experts to refine model insights, ensuring alignment with real business needs.

What You Bring

• You enjoy solving real-world business problems involving data-driven optimization and ML modeling - and have been doing that successfully for a while.
• You are comfortable measuring and optimizing the direct business impact of work.
• You are experienced with SQL, handling large-scale data, and are comfortable with at least one programming language (Python, R, etc.).
• You have experience building ML models using modern ML approaches like Neural Nets or Tree-ensembles from scratch for new applications - making decisions relating to which supervised labels to use, the metric to optimize for, and the features likely to be useful.