How to turn a complex technological environment to real business advantage create a business case?
Working with UNION Insurance, we have developed a a modern, scalable data environment which is not only technologically stable, but also supports the long-term analytics and AI initiatives. The aim was not “just” a new architecture, but creating a secure basis for a data-driven operation.
AI in practice - 5 weeks of tailor-made training
A key element of the project is the transfer of practical knowledge was. An 5 weeks of AI training tailored to your specific needs for specific needs, where the focus was on solutions that could be applied immediately.
Our main themes:
- Azure AI services – from Azure basics to proprietary AI models. Practical Language, Vision, Speech, Translate and Document Intelligence solutions.
- Azure OpenAI – RAG architecture, prompt engineering, model fine-tuning
- Databricks Machine Learning – ML solutions for business environments
- AutoML and AI-based SQL queries – how AI is becoming part of everyday work
The aim was to use AI not as a separate projectbut as a natural tool in day-to-day operations and to be part of the design of new solutions from the outset.
Data synchronisation - when the standard solution is not enough
The core element of technical implementation Synchronisation of 13 critical business tables from the on-premise SQL Server environment to the Databricks cloud platform.
The use of Azure Data Factory was considered, but in the end – with flexibility and cost-effectiveness and flexibility, we opted for a customised approach.
The solution: parameterized Python/PySpark notebooks
The data movement was implemented using custom-built notebooks that:
- Maximum security provide (encrypted access managed in Azure Key Vault)
- Transparently documented (Markdown based structured code)
- Flexible loading logic support:
- MERGE (update + insert)
- INSERT OVERWRITE (full overwrite)
- INSERT INTO (append)
This made it possible for each data type to be atoptimum strategy was applied for each type of data.
Automated operation with Databricks Jobs
A notebookokat Databricks Jobokba have been organized so that:
- supported by the scheduled and manual execution,
- provided the parameterizability,
- can be managed visually in the process dependencies
The result: a stable, automated and well-managed data stream.
More than a technology project
At the end of the project, the result was not just a working technical solution. The UNION team has developed a with internal AI competences that can be exploited in the long term which supports the data-driven decision-making and business innovation.
For us, this cooperation has once again confirmed:
technology becomes truly valuable when we turn it into a business outcome.