Purpose-built visualizations that off-the-shelf BI tools couldn't deliver
A bespoke analytics application that ingests claims, weather, and geospatial data, works out its structure
on its own, and renders interactive dashboards shaped around how the team actually works.
The Challenge
A reinsurance company relied on detailed model output, but the analyses and visualizations they
needed were too specialized to build in Tableau or other BI platforms. They wanted something
purpose-built around their data, and flexible enough to absorb whatever source came next.
Off-the-Shelf BI Fell Short
Tableau and similar platforms were locked to a fixed set of chart types, and couldn't express the analyses the team needed.
Specialized Reinsurance Views
Catastrophe exposure maps and bespoke interactive charts that had to be built from scratch.
Disparate Data Sources
Analyses spanned model outputs, weather feeds, and GIS datasets in different formats and coordinate systems — all of which had to be ingested, reconciled, and explored together.
The Solution
We built the tool from scratch and shaped it around their workflow. It ingests claims, weather, and geospatial data,
works out the data structure on its own, and renders the specialized, interactive views they
couldn't get elsewhere.
Flexible Data Ingestion
The tool takes raw outputs from model data, weather feeds, and shapefiles, and makes them immediately available for analysis, with no per-source setup.
On-Demand Trigger
A cloud trigger kicks off the pipeline on demand, so data is processed only when it's needed.
Ingest Claims, Weather & GIS
Claims data, weather feeds, and shapefiles pulled in as-is.
Python Processing
Python transforms and normalizes each source, handles geospatial data with GeoPandas, and auto-classifies every field.
SQL Server
Processed data persisted to SQL Server, ready to serve.
Interactive Dashboards
A React + TypeScript front end requests data through Python and renders the views.
Triggered on demand in the cloud — point it at a new source and it's explorable immediately, with no schema mapping or manual setup.
Cloud Pipeline, Triggered On-Demand
Runs entirely in the cloud, firing from a trigger only when data needs processing. React requests data from a Python service, which reads from SQL Server.
Automatic Dimensions & Metrics
There's no manual configuration. The app inspects each incoming field and decides whether it's a dimension or a metric, which is what lets it adapt to any dataset on the fly.
Visualizations Built for Reinsurance
Catastrophe exposure maps powered by GIS data using GeoPandas and Shapely, alongside bespoke interactive charts rendered in the browser with full control over layout and interaction.
Results
New Datasets, No Rebuild
Analysts can drop in a new source and explore it the same day, instead of waiting on a new report.
Faster Analysis Turnaround
Work that once took significant manual effort now happens far faster.
Reusable, Customizable Exposure & Weather Visuals
Customizable exposure maps and time-series weather visualizations that analysts can reuse across portfolios and tailor to each question.
