• Responsible AI in hiring and recruitment — assessing the trustworthiness of AI/ML tools used to support hiring decisions, with a particular focus on the fairness and security challenges of LLM-based systems, including how such systems should be audited and governed before deployment. Funded by and in collaboration with Amazon (SoK paper, ongoing).

• Bias and fairness in large language models (LLMs) — bias in an LLM is not spread evenly across the network but concentrated in particular components, so the work first locates which attention heads and feed-forward layers carry it, then fine-tunes only those components against a fairness-aware training objective (ongoing).
• Fair and privacy-preserving distributed machine learning — designed a group-aware federated learning architecture that provides equitable model performance across demographic groups while training under differential privacy. Publication under review (NDSS '27).

Research and development across EU-funded projects including DAIS (EU flagship project, 47 partners) and DADAP (hospital and clinical partners working on mental health disorders), collaborating closely with academic, industrial, and clinical partners to move privacy-preserving machine learning from research into practice.

• GDPR compliance from the “right to be forgotten” — built a data-deletion (machine unlearning) framework that lets individuals remove the influence of their data from a collaboratively trained model in distributed machine learning systems where the central server coordinating training is not a trusted party. The framework enforces the right to be forgotten using encryption, so that each individual can autonomously request deletion without relying on the central server or revealing their request. Publication (CIKM '25).
• Scalable privacy preservation in distributed machine learning — made strong encryption practical at scale in federated learning by compressing client updates before encrypting them, so that encryption cost no longer scales with model size, keeping updates encrypted throughout aggrgetaion while holding communication and computation costs low and model performance intact. Publications (PMCJ '25, Workshop at NeurIPS '24, ISPEC '23).
• Real-world model deployment — built a physical testbed and used it to train and deploy speech emotion recognition models collaboratively across heterogeneous edge devices, measuring efficiency, fairness, and privacy trade-offs under real deployment conditions. Publications (IJCNN '25, FedCSIS '23).
• Healthcare AI on sensitive data — benchmarked a range of machine learning models, including recent transformer architectures, on clinical mental-health tabular records collected in Swedish hospitals, and provided a broad evaluation of predictive performance and of how that performance varies across demographic groups, in order to give clinicians reliable decision support. Publication (PharML at ECML PKDD '26).

• Enterprise software development (ERP) — built new Odoo modules end to end (Python backend, JavaScript front end), supporting finance and supply-chain operations for business clients.
• Systems integration — connected those modules to clients' existing finance and supply-chain systems, keeping data and processes consistent across both.

• Anomaly detection in information networks — developed graph-based methods that flag suspicious nodes by tracking their activity and how it changes over time, an approach applicable to fraud detection, security monitoring, and abuse detection in social networks.