The Client
An international financial institution operating in both corporate and retail banking, serving several million clients in the CEE region. The organization had a mature data warehouse environment and a well-developed analytics team but was still at an early stage of implementing AI and GenAI solutions in decision-making processes.
The problem
The management board expected the implementation of AI solutions supporting risk assessment, automation of operational processes, and analysis of client documents. The internal data science team focused mainly on experimental models, while the organization lacked the capabilities required to build stable production pipelines and integrate AI solutions with core banking systems.
The key challenges
- lack of end-to-end ownership for the technical implementation of ML/GenAI solutions,
- difficulties transitioning from proof of concept (POC) to a production environment,
- integration of more than 10 data sources (core banking, CRM, scoring systems, PDF documents, transactional systems),
- strict regulatory and audit requirements.
The organization needed an experienced consultant who could take responsibility for the technical delivery of AI solutions in a project-based model.
The solution
Within a consulting model, we engaged a Senior AI Engineer as an individual contributor responsible for the full implementation lifecycle.
The project scope included:
- designing and implementing end-to-end data pipelines using SQL and Python (NumPy, PyTorch),
- building workflows supporting scoring models and document analysis using LLMs,
- designing a GenAI architecture supporting analysis of contracts and credit applications,
- designing ontologies and intermediate layers connecting data from multiple systems,
- conducting technical workshops with IT and risk teams,
- transitioning from POC to a stable production deployment with model monitoring.
The consultant worked within a small, autonomous team, collaborating directly with product leaders and client architects.
The conclusions
- 60% reduction in the time required to analyze credit documentation through the use of GenAI.
- Implementation of production ML pipelines compliant with audit and regulatory requirements.
- Improved decision-making processes through integration of models with core systems.
- Creation of an architectural blueprint for future AI projects within the organization.
The client obtained a functioning and scalable AI solution supporting key business processes, along with a practical model for delivering AI projects from analysis to production.
