Automated Data Extraction from Product Labels
Following the success of Nexus AI, MSM's Generative AI department identified the need to industrialize its processes and extend the technology to new areas involving unstructured data. The challenge lay in automating manual and repetitive tasks within two complex workflows: reading nutritional information from labels with varied formats and analyzing extensive public tender documents.
WhiteBox provided an "AI Engineering Implant" service, integrating an MLOps and Generative AI expert directly into the client's team. Pipelines were developed on Microsoft Azure combining OCR and LLMs for two specific use cases: a system to extract and normalize data from product label photos, and a tool to identify strategic entities (such as budgets, deadlines, and requirements) in tender specifications.
The collaboration enabled the industrialization of AI models, ensuring a scalable, efficient, and GDPR-compliant infrastructure. This freed up teams from low-value added tasks, accelerated the preparation of tender offers, and guaranteed reliable and standardized data extraction, consolidating Generative AI as a transversal and strategic tool for the company.
Generative AI for Textile Design
