Skipt to contentSkipt to footer

Reference

Efficient document processing using AI solutions

Our customer:

North Data

North Data is the number 1 platform for company information! If you are searching for information – such as balance sheets – on any company, you will most likely use northdata.de.

Project goal and challenge

The goal was the automatic extraction of key figures such as profits or personnel expenses from millions of unstructured annual reports of British companies, most of which are available as scanned PDFs, for example as images. These reports are very inconsistent. This made data extraction and processing using methods previously available on the market inefficient and cost intensive.

An illustrative example of data extraction using AI. On the left is the input: unstructured pages from an annual report containing text, tables and diagrams; on the right is the output: JSON code containing the key figures sought.

Solution and measures

In collaboration with North Data, cronn developed an innovative solution using modern AI technologies, in particular Large Language Models (LLMs). The following measures were carried out:

  • Data analysis: Development of a representative test dataset from millions of PDFs and creation of a ground truth to validate the data processing results.
  • **Technology evaluation:**Selection and evaluation of various AI models (Google Gemini, ChatGPT, Mistral) and OCR technologies to determine their effectiveness in recognising and extracting texts and tables from the documents.
  • Development of a machine learning pipeine: In-house development of a Java-based solution for classifying and structuring PDF documents in order to optimize the amount of data for the AI models.
  • **Integration into the production system:**Fast and seamless implementation of our solution into North Data’s Java-based backend, including automated quality controls (MLOps).
cronn reference quote northdata

We have been working successfully with cronn for several years now. In this project too, they quickly understood our requirements. Together, we were able to design an AI-based solution and put it into production. We value cronn’s expertise in the areas of AI and backend and are very satisfied with the implementation.

Frank Felix DebatinCEO of North Data

Customer benefit

cronn supported North Data in expanding its existing system to integrate AI-based technologies into extracting key figures from UK annual reports. As a result, North Data now has a powerful, accurate, and cost-effective solution for automated processing of copious amounts of data. It is particularly noteworthy that the attractive price per processed annual report makes this solution not only technically efficient, but also economically advantageous.

Other projects