Applied ML Engineer/Data Scientist

Your role

Our data and tech give us an edge in risk assessment, pricing and premium calculation. This role turns that into accurate models that support our carrier partners and our growth, builds tuned models where they beat rules and off-the-shelf LLMs in claims (triage, intervention, return-to-work, fraud), and responsibly explores how data can help employees stay healthy through grape Care.

We’re looking for an applied ML engineer / data scientist to build predictive models from the ground up on real, regulated insurance data. You’ll own problems end to end, from effective-dated records to production, in a domain where correctness and responsible use genuinely matter.

Starting points include pricing and risk, portfolio monitoring, claims and payment patterns, triage and return-to-work, absence, and fraud detection. grape sits on a rich dataset, though, and you’ll help find the next high-value use cases. You’re candid about what the data can support, and you propose simpler approaches when they’re the right answer.

This is not an LLM role. LLM and agent workflows and document extraction sit with our engineering teams, and you’ll use their outputs as features.

What you'll do:

Build and validate models

• Own the full model lifecycle: feature engineering, training, evaluation, explainability, deployment, and drift monitoring.
• Backtest on the historical book and run models in shadow mode alongside human decision-makers, reporting where and why they disagree.
• Produce audit-ready validation and governance documentation that actuarial, compliance, and carrier partners can review.

Platform and MLOps

• Deploy models on grape’s cloud platform within strict security and data-residency requirements for health data.
• Set up experiment tracking, model versioning, and production monitoring, partnering with engineering on integration.

Shape the problem space

• Identify and prioritize high-value use cases as you explore the dataset, rather than waiting to be handed problems.
• Turn ambiguous business questions into well-defined, measurable modelling.

Responsible & explainable by design

• Build privacy, fairness, and explainability into every model that touches personal or health data, working with compliance and legal on nDSG/GDPR boundaries.
• Models support decisions; accountable owners make them. Pricing, reserving, and claims work requires licensed-actuary sign-off, and any automation requires governance approval first.

Collaborate

• Explain model behavior and limits to underwriting, claims, actuarial, and compliance teams.

What you'll need:

• Strong applied ML and statistics: supervised and unsupervised learning, anomaly detection, time-series, and survival/duration modeling.
• Proven experience building models on real-world, relational, effective-dated data, including imbalanced outcomes.
• Fluent Python (pandas, scikit-learn, a modern ML stack) and strong SQL.
• Solid grasp of validation, calibration, and explainability (e.g. SHAP).
• Production MLOps: experiment tracking, versioning, and drift monitoring.
• Sound judgment on data privacy and modelling.
• Fluent English.
• Bias for action, ownership, and clear cross-functional communication.

Big plus:

• Insurance, actuarial, or risk modeling (pricing, reserving, loss ratios).
• Health, disability, or absence data, especially Swiss insurance (KTG, UVG); fraud detection in claims.
• Regulated special-category data under nDSG/GDPR, or model risk management.
• Google Cloud, including security-restricted deployments.
• German language skills.