Senior Engineer, Machine Learning - Virta Health

Virta Health is on a mission to transform type 2 diabetes and weight-loss care by helping people reverse their metabolic condition through innovations in technology, personalized nutrition, and virtual care delivery. We have raised over $350 million and partner with health plans, employers, and government organizations. Join us on our mission to reverse diabetes and obesity in one billion people.

About the Role

As a Senior Engineer, Machine Learning on Virta's Applied AI team, you build the models that shape how Virta shows up for each member: who needs a coach's attention this week, which nudge will land, what moves someone closer to reversing their disease. You take a fuzzy clinical or operational question, sharpen it into something a model can actually answer, and carry it the whole way to production.

You are hard to fool. A number that looks good on a slide doesn't convince you until you've checked what's underneath it: whether a leak inflated it, or whether it's quietly wrong for a segment it doesn't call out. You start simple, prove the baseline honestly, and add complexity only once it earns its place.

Your models don't just ship, they stay healthy without you babysitting them. As Applied AI moves from linear to exponential, you build models on Virta's deep, longitudinal member data and put them on the ML platform that retrains them, watches their quality, and pulls you in only when something is genuinely wrong. What one engineer can deliver stops being capped by how many models they can keep alive.

None of this happens solo. You hand models to product engineering to get them in front of members and care teams, lean on ML Ops's infrastructure to keep models running, sit with clinicians to pressure-test what the data is really saying, and work with Product Analytics to design experiments sharp enough to trust, and to separate what actually drives better outcomes from what only correlates with them.

Responsibilities

• Problem Framing & Modeling: Turn ambiguous clinical and operational questions into well-scoped ML tasks, and build strong baselines before adding complexity.
• Model Delivery at Scale: Build predictive models that drive member care, like who is at risk of disengaging, who benefits most from a coach's outreach, or where a member's health is trending. Partner with product engineering to put that signal in front of members and care teams.
• Self-Maintaining Models: On top of the ML platform, build the per-model logic that keeps each model healthy in production (retraining triggers, drift and anomaly detection, continuous evaluation), with compliance considerations built in from the start.
• Evaluation & Error Analysis: Design offline and online evaluations, choose metrics suited to imbalanced, delayed, and noisy healthcare outcomes, and run careful error analysis on model failures.
• Causal ML & Experimentation: Build causal models validated against real clinical experiments, and partner with Product Analytics on variance-reduction techniques (CUPED/CUPAC) to sharpen experiment sensitivity on Virta's member history.
• Signal Integration: Build ML-and GenAI-derived signals that product squads can consume to guide personalized coaching and member experiences.
• Technical Leadership: Lead complex ML initiatives from problem framing through production, coordinating technical decisions and dependencies across Applied AI, ML Ops, Product Engineering, and Product Analytics. Share knowledge and practices that strengthen model quality and delivery across the team.

Must Haves

• 5+ years building, evaluating, and deploying production-grade Machine Learning systems for real-world use cases.
• Strong foundation in traditional ML (predictive modeling, tree-based models like XGBoost, time-series analysis on structured longitudinal data).
• Rigorous evaluation and experimental design: statistical reasoning, offline/online evaluation, and a habit of identifying leakage, selection bias, and confounding in messy real-world data.
• Ability to explain model behavior, tradeoffs, uncertainty, and limitations clearly to both technical and non-technical stakeholders.
• Experience designing and analyzing experiments to determine what actually drives an outcome, not just what correlates with it.
• Depth in one or more of: causal ML and experimentation, recommendation/ranking systems, predictive modeling, or NLP/LLM-based systems.
• Responsible use of AI-assisted development to meaningfully accelerate and improve model development.

Nice to Haves

• Background in clinical or health-tech ML (claims, EHR, clinical notes, scheduling, utilization, or member engagement data), and experience with PHI, HIPAA-aware systems, or other regulated data.
• Judgment about when traditional ML outperforms LLMs, and when an LLM is the right tool.