Data Scientist Principal, AI Development and Governance

General Dynamics Information Technology•USA TN Home Office (TNHOME), TN

3d•$119,000 - $161,000•Remote

About The Position

Own your opportunity to turn data into measurable outcomes for our customers’ most complex challenges. As a Data Scientist Principal at GDIT, you’ll power innovation to drive mission impact and grow your expertise to power your career forward. This role exists to turn a multi-billion-record, multi-payer healthcare claims warehouse, the Healthcare Fraud Prevention Partnership (HFPP) Trusted Third Party (TTP), into fraud, waste, and abuse (FWA) findings trusted enough for Partners and investigators to act on. The team's models and tooling increasingly depend on machine learning and generative AI, and this is the role that owns what "trustworthy" means for both. Roughly half your time goes to building models, the other half to setting the standards the rest of the Data Science team builds against. This is a senior individual-contributor role with no direct reports.

Requirements

Master's degree in a quantitative field (statistics, computer science, engineering, applied mathematics, economics, or related), or a Bachelor's in one of those fields with equivalent hands-on experience.
8+ years building, validating, and deploying ML models on real-world data, including a track record of setting technical standards that other data scientists work against.
Working knowledge of responsible-AI and model-risk practice: documentation, monitoring, bias and drift detection, and what production-ready governance looks like for a model whose output drives decisions about providers.
Experience evaluating generative AI and LLM use cases for both feasibility and risk, including cases where your answer was that an LLM shouldn't be used yet.
Competence in Python and SQL, including feature engineering inside a data warehouse at very large scale.
2+ years working with healthcare claims data (Medicare, Medicaid, or commercial), plus working knowledge of medical terminology and healthcare coding systems (ICD-10, CPT, HCPCS, DRG).
Experience presenting technical and governance decisions to clients, partners, or auditors. You should be able to defend a methodology choice to a technical reviewer and explain that same decision to someone who isn't one.

Nice To Haves

Prior experience in a formal model-risk or responsible-AI role, even outside healthcare.
Graph or network analytics, entity resolution, or record linkage.
Experience piloting generative AI tools in a regulated or high-scrutiny setting.
AWS and/or Snowflake environments, including Snowpark or model lifecycle tooling.
Experience with payer coverage policy (LCDs, NCDs, private carrier policies) and industry claim edits (NCCI).

Responsibilities

Set the modeling and validation standards the Data Science team works against; how models get documented, monitored, and checked for drift and bias. You'll review the team's models against that bar before they go to production and recommend what must change first.
Write and maintain the program's responsible-AI and GenAI policy. The harder half is generative AI inside the FWA pipeline itself, like case narrative summarization or investigator-facing drafts, where a weak output lands in front of an investigator. Internal tooling such as code assistants needs a policy too, and it's the easier one to write.
Build and ship FWA models yourself. Supervised risk scoring against the claims warehouse, feature engineering at claim-record scale, and validation under heavy class imbalance and fraud schemes that shift faster than confirmation arrives.
Walk HFPP Partners and internal auditors through how a given model or AI-assisted step works, including validation results and controls. Expect to defend methodology choices to people whose job is finding the gaps in them.
Decide what's worth piloting as generative AI capability shifts and say no to what isn't ready for a healthcare FWA context yet.
Work across a multi-disciplinary team of Data Scientists, BI Developers, and FWA Subject Matter Experts (100% remote, distributed across the US). Governance questions come to you regardless of which sub-team raised them.