European Life Technologies Hungary is working to make liquid biopsy testing more widely available and to use artificial intelligence to support the analysis of laboratory data. At Portfolio Deep Tech 2026, company leaders described how the technology can help researchers and specialists process information needed to detect and monitor cancer.
Orbán Előd, co-founder and chief executive of European Life Technologies Hungary and managing director of Enterprise Group Technologies, said the company aims to become a major international player in liquid biopsy and make testing as widely available as diagnostic imaging. In some cases, he said, relevant information about a breast cancer recurrence can be obtained from blood or urine 6–9 months before imaging tests.
According to the company, international expansion requires more than additional staff: it also calls for technology that multiplies specialists’ capacity. It is therefore approaching AI as a tool, a process and a system for supporting organizational operations.
From research knowledge to variant assessment
Pintér Lajos, chief operating officer of European Life Technologies Hungary, who leads both software development and biology teams, presented three practical applications. The first is intended to preserve research knowledge that could otherwise be lost when employees leave. Researchers spend 5–10 minutes a day recording their work and experimental results. The AI processes those records alongside earlier protocols and the project description, produces standardized documentation and suggests further research directions. The researcher checks and approves the content in every case.
The second system prepares assessments of genetic variants. Evaluating a single variant requires a review of scientific literature, clinical findings, treatment options and other evidence, while one test can reveal hundreds of variants. The system gathers background information and prepares a classification using a versioned set of rules.
A separate program, rather than the language model, calculates the scores. Another AI then checks references and whether claims are supported, while a specialist gives final approval. Pintér said the acceptance rate for prepared materials exceeds 90%: that measures expert acceptance, not diagnostic accuracy. Reasons for rejection are fed back into the system to support future assessments.
Querying data and protecting privacy
The third development is a natural-language query layer built on the company’s existing bioinformatics and laboratory workflow management systems. It allows researchers and analysts to look for connections among previous cases, variants, therapies and clinical trials without commissioning separate development work. A use case intended to help doctors interpret data about their own patients, however, is not yet fully operational.
Under the data protection approach presented, personal information is masked before data is sent to cloud-based models. A locally run language model isolated from the internet helps process free text.
Pintér said the difficult-to-replicate asset is not AI alone, but a knowledge base built over several years that handles more than 20 connected data types. In this division of labor, the machine reads, writes and prepares material, while a person makes the decision. Orbán added that, alongside technology, a competitive advantage requires a corporate culture that is open to learning, collaborative and capable of rapid experimentation.



