Decision Support System for Underwriters
Development of an AI-based Proof of Concept (PoC) to support underwriters in quote evaluation by providing confidence scores derived from structured data.
Reinsurance Client
2024
Know More
The client is one of the world’s largest reinsurance companies, known for driving innovation through digital services tailored to primary insurers. Their B2B platform enables reinsurance agents and underwriters to manage, price, and reinsure complex insurance products, with digital certificates issued post-submission.
Problem
Manual Quote Decisions and Limited Decision Support Capabilities
Underwriters needed to assess insurance quote submissions across a range of products and project types. This process was largely manual and lacked systematic guidance on whether a quote should be accepted or declined. Although the client maintained relevant submission data across various systems, this information had not been leveraged for predictive modeling or decision support, creating inefficiencies and inconsistency in evaluations.
Solution
End-to-End AI Prototype for Risk Scoring within Dataiku
An AI-based prototype was developed within the client's existing Dataiku environment to assist underwriters in evaluating submissions. The solution included data ingestion from multiple Azure-hosted Microsoft SQL databases, data transformation using Python scripts, and exploratory analysis to understand relevant patterns and features. A machine learning model was then built and trained to predict quote decisions with a confidence score, helping underwriters make more informed and consistent decisions. The PoC covered the full ML lifecycle – from data preparation and feature engineering to model evaluation and interpretation.
Technologies & Frameworks
AI-Driven Decision Support with Dataiku, Azure, and Python
Dataiku – served as the central data science platform for managing the full AI workflow (data prep, model training, evaluation)
Azure MS SQL – primary source system for historical quote data used to train and validate the model
Python (pandas, scikit-learn, XGBoost) – used for data transformation, model development, feature analysis, and hyperparameter tuning
Microsoft Azure – hosted the client’s infrastructure and data, ensuring compliance and scalability
Jupyter Notebooks – enabled prototyping and in-depth performance analysis during the PoC phase
More Works
Decision Support System for Underwriters
Development of an AI-based Proof of Concept (PoC) to support underwriters in quote evaluation by providing confidence scores derived from structured data.
Reinsurance Client
2024
Know More
The client is one of the world’s largest reinsurance companies, known for driving innovation through digital services tailored to primary insurers. Their B2B platform enables reinsurance agents and underwriters to manage, price, and reinsure complex insurance products, with digital certificates issued post-submission.
Problem
Manual Quote Decisions and Limited Decision Support Capabilities
Underwriters needed to assess insurance quote submissions across a range of products and project types. This process was largely manual and lacked systematic guidance on whether a quote should be accepted or declined. Although the client maintained relevant submission data across various systems, this information had not been leveraged for predictive modeling or decision support, creating inefficiencies and inconsistency in evaluations.
Solution
End-to-End AI Prototype for Risk Scoring within Dataiku
An AI-based prototype was developed within the client's existing Dataiku environment to assist underwriters in evaluating submissions. The solution included data ingestion from multiple Azure-hosted Microsoft SQL databases, data transformation using Python scripts, and exploratory analysis to understand relevant patterns and features. A machine learning model was then built and trained to predict quote decisions with a confidence score, helping underwriters make more informed and consistent decisions. The PoC covered the full ML lifecycle – from data preparation and feature engineering to model evaluation and interpretation.
Technologies & Frameworks
AI-Driven Decision Support with Dataiku, Azure, and Python
Dataiku – served as the central data science platform for managing the full AI workflow (data prep, model training, evaluation)
Azure MS SQL – primary source system for historical quote data used to train and validate the model
Python (pandas, scikit-learn, XGBoost) – used for data transformation, model development, feature analysis, and hyperparameter tuning
Microsoft Azure – hosted the client’s infrastructure and data, ensuring compliance and scalability
Jupyter Notebooks – enabled prototyping and in-depth performance analysis during the PoC phase
More Works
Decision Support System for Underwriters
Development of an AI-based Proof of Concept (PoC) to support underwriters in quote evaluation by providing confidence scores derived from structured data.
Reinsurance Client
2024
Know More
The client is one of the world’s largest reinsurance companies, known for driving innovation through digital services tailored to primary insurers. Their B2B platform enables reinsurance agents and underwriters to manage, price, and reinsure complex insurance products, with digital certificates issued post-submission.
Problem
Manual Quote Decisions and Limited Decision Support Capabilities
Underwriters needed to assess insurance quote submissions across a range of products and project types. This process was largely manual and lacked systematic guidance on whether a quote should be accepted or declined. Although the client maintained relevant submission data across various systems, this information had not been leveraged for predictive modeling or decision support, creating inefficiencies and inconsistency in evaluations.
Solution
End-to-End AI Prototype for Risk Scoring within Dataiku
An AI-based prototype was developed within the client's existing Dataiku environment to assist underwriters in evaluating submissions. The solution included data ingestion from multiple Azure-hosted Microsoft SQL databases, data transformation using Python scripts, and exploratory analysis to understand relevant patterns and features. A machine learning model was then built and trained to predict quote decisions with a confidence score, helping underwriters make more informed and consistent decisions. The PoC covered the full ML lifecycle – from data preparation and feature engineering to model evaluation and interpretation.
Technologies & Frameworks
AI-Driven Decision Support with Dataiku, Azure, and Python
Dataiku – served as the central data science platform for managing the full AI workflow (data prep, model training, evaluation)
Azure MS SQL – primary source system for historical quote data used to train and validate the model
Python (pandas, scikit-learn, XGBoost) – used for data transformation, model development, feature analysis, and hyperparameter tuning
Microsoft Azure – hosted the client’s infrastructure and data, ensuring compliance and scalability
Jupyter Notebooks – enabled prototyping and in-depth performance analysis during the PoC phase
More Works