ARTIFICIAL INTELLIGENCE WITH SENSITIVE DATA

Sovereign AI solutions from ULX

The artificial intelligence (AI) revolution has ushered in a new era of corporate efficiency, but at the same time, it has raised critical questions about data security and digital sovereignty. At ULX, we recognize that the true power of AI—at least for enterprise use—lies not in the use of various public generative AI services (e.g., ChatGPT), but in companies' own internal knowledge bases.

Our goal is to enable our customers to use the latest predictive and generative AI technologies without their valuable and sensitive data leaving the secure infrastructure of their organization.

The principle of data sovereignty: Why keep AI in-house?

When using most public AI services, uploaded data—whether contracts, financial data, or internal processes—is removed from the organization's control and often used to further train the service provider's models. In contrast, the approach proposed by ULX follows the principle of "Privacy by Design":

  • Protection of sensitive data: Documents, invoices, customer data, and internal knowledge base information are processed exclusively within companies' internal, protected IT environments.

  • Local model training: Models are fine-tuned and trained on our own infrastructure, so the business intelligence used remains the exclusive intellectual property of the company.

  • Compliance: We help you comply with strict data protection regulations (GDPR, banking secrecy, commercially confidential information) without having to sacrifice the benefits and opportunities offered by AI.

Technology foundation: Red Hat OpenShift AI

Our AI strategy centers on the Red Hat AI portfolio, primarily the Red Hat OpenShift AI platform. This solution provides a unified, scalable, and secure environment for data scientists and DevOps engineers to manage the entire model lifecycle.

What do we use OpenShift AI for?

  • Model development and prototype production

  • Data processing workflows, building automated pipelines from data cleaning to model validation

  • OpenShift AI Inference Server enables highly available, scalable serving of trained models

Our competencies: Predictive and Generative AI applications

At ULX, we don't just build platforms, we also provide support and assistance in implementing specific business use cases.

Predictive AI (forecast-focused) solutions

With predictive AI, it is possible to predict future events based on past data (for example, by recognizing patterns):

  • Financial anomaly detection: Preventing fraud and identifying suspicious transactions.

  • Maintenance prediction: Prediction of industrial system failures based on sensor data.

  • Customer churn prediction: Identifying customers who are about to leave before they make their decision.

Generative AI and RAG:

At ULX, we don't primarily use generative AI for chatting. With the help of RAG (Retrieval-Augmented Generation) technology, we connect large language models (LLMs) with the company's own knowledge bases and document repositories to implement services such as:

  • Intelligent document processing: Automatic extraction and structuring of contracts and invoices.

  • Internal knowledge base assistant: Creating AI bots that respond to employee questions based solely on the company's internal policies, technical specifications, and previous cases.

  • Code generation and IT automation: Increase developer efficiency with internal, secure code assistants.

Our services throughout the AI lifecycle

AI is a complex field waiting to be explored, which interests everyone but is easy to get lost in. At ULX, we have built up expertise over the past few years that enables us to help our clients from the very beginning to the successful completion of their projects:

  • AI Readiness Assessment: We assess the company's data assets and infrastructure suitability (GPU requirements, storage capacity) and identify areas where AI solutions can be most effective.
  • Infrastructure planning and GPU optimization: We develop the hardware and software foundations necessary for running AI (GPU integration, OpenShift AI implementation).
  • Data preparation and model training: We help with the anonymization and cleaning of sensitive data and the local (on-premise) training of models.
  • Inference and integration: We integrate the finished models into existing enterprise applications (e.g., ERP, CRM, workflow systems) via APIs.
  • Continuous monitoring and fine-tuning: We monitor the accuracy of the models (drift detection) and, if necessary, retrain them with the latest data.

Why choose ULX when you need a secure and sovereign AI solution?

  • 25 years of experience with large enterprises: We understand the strict requirements of the banking, insurance, and government sectors.

  • Purely open source: Our solutions are transparent, auditable, and free from secret, unverifiable algorithms.

  • Holistic approach: AI is not an isolated island at our company, but an integral part of the OpenShift-based cloud-native ecosystem.

  • Local expertise: We provide Hungarian-language consulting and support to assist with model implementation and operation.

Want to leverage the power of AI without compromising your data security?

Contact us and request a presentation of our AI strategy or the development of a Proof of Concept (PoC) project, where we will demonstrate the power of the technology using your own data in your own environment!