We conducted end-to-end AI consulting to identify high-value use cases and deploy GenAI/LLM capabilities.

• Challenge: There was no clear roadmap for where GenAI and LLMs could deliver measurable business value. Process descriptions were fragmented, limiting automation and control.
• Approach: Performed stakeholder workshops, value-mapping and feasibility scoring to prioritise AI opportunities. Mapped critical workflows and created a standardised, AI-ready process description format. Development of proof-of-concept prompts and micro-pipelines for LLM-supported standardised process documentations.
• Outcome: A prioritised GenAI roadmap (covering the four most important use cases) with estimated ROI and implementation plan; prototype workflows for LLMs designed for knowledge extraction and decision support, ready for pilot operation. Implementation of an intelligent process management system.

We helped an industrial engineering client convert business needs into production-ready AI features and led a successful Retrieval-Augmented Generation (RAG) rollout.

• Challenge: IT access to technical design documents was incomplete and inconsistent. However, the engineering teams required high-performance and reliable access to design documents, including test reports and manufacturers manuals.
• Approach: Gathering user requirements, developing pipelines for document capture and embedding. Designing a retrieval layer for the vector storage and developing prompt templates. End-to-end project management: backlog management, sprint planning, model selection, performance metrics and coordination of deployment.
• Outcome: A production-ready RAG application has been deployed, which provides accurate, context-aware answers from technical documents, thereby reducing document search times and shortening review cycles. Monitoring dashboards for model drift and relevance have been made available.

We developed a mathematical credit risk portfolio model and a working prototype to measure and manage credit portfolio risk in energy trading.

• Challenge: The client required a robust, quantitative model to assess the credit risk of its energy trading portfolio. In particular, the aim is to determine the Credit Value at Risk (Credit VaR) loss distribution function. In doing so, correlations, diversification and granularity must be taken into account appropriately. A prototype has to be developed on this basis.
• Approach: Development of the model architecture for a Monte Carlo simulation of credit defaults. This includes a model for credit default events, the credit rating transition matrices and the aggregation of scenarios. Calibration of the model parameters using market data. Implementation of a prototype of the model in code, which includes components for analysing the Credit VaR loss distribution function. Creation of visualisation and export functions for reports.
• Outcome: A fully functional prototype of a validated Credit VaR portfolio model based on a Monte Carlo simulation of credit default events. This includes the capability to perform sensitivity analyses and stress test scenarios. Credit risk managers are thus enabled to make informed decisions regarding portfolio management, risk hedging and capital allocation.