Automatic damage detection after natural disasters

Settlement within 10 days instead of 9 months due to automatic damage detection

Challenge

After natural disasters, an insurance company wants to quickly and cost-effectively assess the damage to its insured buildings. This requires an infrastructure that can process a large amount of image material in a timely manner. In addition, a model must be developed that allows the prediction for different building types and regions.

Solution

Various machine learning models are developed, evaluated and made available. These are the basis for automatic damage detection. A pipeline is built in the Amazon AWS Cloud that seamlessly integrates both the model training and the productive use of the results. Due to the complete integration of all processes in the AWS infrastructure, the product is scalable and can be automatically adapted to the current demand in a cost-efficient manner during idle and load times.

Result

The models can be trained and adapted in a scalable pipeline. Via an API, the results can be used for a wide range of applications for automatic damage detection. Damage classification makes it easier to estimate payout amounts and initiate payout processes automatically.

Challenge

An automotive company would like to visualise various market-specific data in order to create a Competitive analysis for the US market.

Solution

There will be a interactive and Flexible application, including of different maps with two different views implemented.

Result

Relevant markets are identifies, analyses and visualises. The dealer or the respective sales department have the possibility to compare the direct competition with their own product and to visualise the relevant data.

Risk assessment in insurance applications

Challenge

An international insurance group wants to offer an online-compatible building insurance tariff. In order to optimise the application process, as few risk questions as possible should be asked. To ensure that the risks can still be assessed correctly, the risk questions are to be replaced by data from external providers.

Solution

About 1,300 characteristics from external providers are assessed. From these, targeted candidates are selected to replace the risk questions. With the help of a substitution algorithm, each risk question is checked and, if necessary, substituted by external data. The close cooperation with the actuaries ensures a professionally sensible pricing.

Result

78 % of the previous risk questions can be preassigned, replaced or omitted with the same forecast quality. New external data can be checked and integrated by the actuaries themselves at any time.

Challenge

An automotive company would like to visualise various market-specific data in order to create a Competitive analysis for the US market.

Solution

There will be a interactive and Flexible application, including of different maps with two different views implemented.

Result

Relevant markets are identifies, analyses and visualises. The dealer or the respective sales department have the possibility to compare the direct competition with their own product and to visualise the relevant data.