Head of Data & AI

About Cabrella
Cabrella insures the parcels most carriers would rather not touch — loose diamonds and finished jewelry, precious metals, graded trading cards, fine art in transit — using software we build ourselves: a native TMS with coverage inside the shipping workflow, an insurance API partners embed in their own flows, pre-transit risk scoring, and a white-labeled claims platform.
Coverage runs to $150,000 per parcel and $1,000,000 per freight shipment, across 118 countries, on A+ rated admitted paper, with a 90% claims approval rate.
All of it throws off data: quotes and binds, declared values by commodity class, carrier scan and exception feeds, claims through payout, subrogation recoveries, partner API traffic. Today it informs decisions. It does not yet run them.
About the Role
We are hiring a senior leader to build data and AI as a company-wide capability, with a seat at the leadership table, a budget, and ownership of the vendor stack. You will set the strategy, stand up the platform and governance behind it, ship the BI and conversational layer that gets it used, and put models into production where they change an underwriting, pricing, claims, or fraud decision — accountable for what that does to loss ratio and margin.
You have managed distributed or outsourced delivery teams: clear specs, async review, remote quality control, and vendor scope ownership.
What You Will Own
• Data strategy and architecture. Target-state platform and the roadmap to it: ingestion from the TMS, policy, claims, carrier feeds and partner APIs; modeling, lineage, and one definition of every metric the company argues about.
• Business intelligence and self-serve analytics. A governed Power BI estates that leadership, underwriting, claims, and account management rely on metrics they trust.
• Applied AI and automation. Conversational access to our data, document extraction across claims intake, and automation of the manual steps in quoting, exception handling, and adjudication.
• Predictive and fraud models. Claims fraud detection, loss propensity and severity by lane and commodity, pricing and coverage-uptake analytics, and churn signals across the book.
• Governance and model risk. Privacy, access control, retention, auditability, and clear documentation of how our models make decisions.
• Team, partners, and budget. Directing the analysts, data engineers, and ML specialists on the work, whether internal or partner-side; owning data platform and AI spend and defending it on return.
Essential Experience
All four areas below are required. Candidates without hands-on depth in each will not advance. Be ready to walk through systems you built and what they were worth.
1. Modern data strategy, set and delivered
• You have owned data strategy for a company or major business unit — written it, sold it to an executive team or board, funded it, delivered against it.
• Built or migrated a modern stack: cloud warehouse or lakehouse, managed ingestion, transformation as version-controlled code, orchestration, semantic layer with governance.
• Strongly preferred: shipping, logistics, parcel, supply chain or insurance. You know carrier and tracking data, policy and claims models, loss ratio, frequency and severity, reserving, and subrogation — and how dirty this data is in the wild.
2. Power BI at strategy scale, with automation
• Deep, current Power BI ownership: semantic models and star schemas, DAX, row- and object-level security, incremental refresh, deployment pipelines, capacity management, and embedded analytics for internal customers.
• You have set BI strategy, not just built reports: certified datasets, a metric layer people trust, report rationalization, and a self-serve model that cut ad-hoc requests instead of multiplying dashboards.
• Workflow automation alongside it — Power Automate, Microsoft Fabric, Python services — that removed named hours of manual work, with the before-and-after numbers.
3. Conversational AI over data, and automated anomaly detection
• You have put a working “talk to your data” interface in front of business users — Claude, Copilot for Power BI, OpenAI, or equivalent — against a governed semantic model and can explain what made the answers correct: metric definitions, text-to-SQL or DAX generation, retrieval, row-level security at query time, and the questions it should refuse.
• You evaluated it like a product, not a demo: a test set of real business questions, measured accuracy, regression and hallucination checks before release.
• You have shipped automated anomaly detection on data and business metrics — pipeline freshness, volume and schema drift, unexpected movement in premium, claim frequency, severity, or partner API traffic — with alerts routed to an owner.
4. Fraud and predictive models in production
• You have carried out fraud and abuse detection that changed how a business decided — ideally claims, payments, chargebacks, or shipment loss and theft — and can defend the precision and recall trade-off you chose.
• Equally credible on prediction: forecasting, loss and severity, propensity, pricing and elasticity.
• You know what keeps a model useful after launch: feedback from the people acting on its scores, basic monitoring, and being able to explain a result. You judge success in dollars, not accuracy.
Preferred Qualifications
• 10+ years in data, analytics, or AI; 5+ leading teams; at least one stint as the most senior data person in the room.
• Fluent SQL and Python — you can review a model, a pipeline, and a query plan.
• Microsoft data stack — SQL Server or Azure SQL, with Fabric a plus — though equivalent depth on Snowflake, BigQuery, or Databricks transfers. Production experience with LLM APIs, including cost control.
• Comfortable in regulated insurance work: PII and payment data, audit trails, carrier and reinsurer reporting, model documentation.
• Judgment about when to use AI and when not to. We would rather ship three models that hold up than thirty that need explaining away.
If interested, please email your resume to Careers@Cabrella.com

Work arrangement
No