Olivera Kotevska, PhD

Olivera Kotevska, PhD

R&D 100 Award winner. DOE PI, privacy-preserving federated learning & AI safety for science. $33M+ portfolio.

About

Leader in Trustworthy Scientific AI · Computer Science and Mathematics Division · Oak Ridge National Laboratory

She defines the privacy and security foundations for multi-institutional scientific AI — the systems where national labs, universities, and global partners must collaborate without trusting each other.

$33M+
Research portfolio
16
Funded awards
R&D 100
2025 Award winner
84
Publications
15
PhD students mentored
13
Workshops organized

Dr. Olivera Kotevska is a leader in trustworthy scientific AI in the Computer Science and Mathematics Division (CSMD) at Oak Ridge National Laboratory (ORNL), where she directs ORNL’s role in a multi-institutional DOE program on privacy-preserving federated learning for scientific foundation models and shapes AI safety and security for the DOE Genesis Mission. Her research defines the emerging field of trustworthy AI for multi-institutional science, spanning differential privacy, federated learning, gradient privacy, and autonomous scientific computing. She is the 2025 R&D 100 Award winner for PRESTO (Privacy REcommendation and SecuriTy Optimization), a privacy mechanism recommendation system for federated learning at scale, and the recipient of the 2022 Highly Cited Research Paper Award from Applied Energy. Her CVPR 2026 paper on class-level unlearning in vision models was selected as a Highlight.

Prior to joining ORNL in 2019, Dr. Kotevska was an international guest researcher at the National Institute of Standards and Technology (NIST), Maryland, USA, where she was part of the NIST Smart Cities Framework Team and contributed to some of the earliest foundational work in that domain. Before her PhD, she built and shipped production software actively used by millions of consumers — at Nuance Communications (UK, voice and AI systems), Vivo (Brazil, mobile telecommunications), T-Mobile (Macedonia, mobile services), and Renault (France, automotive software). She received her Ph.D. in Computer Science from the Université Grenoble Alpes, France, and B.S. and M.S. degrees in Computer Science and Engineering from the University of Ss. Cyril and Methodius, Skopje, Macedonia.

With over $33M in competitive funding secured across DOE, NNSA, VA, and DoD programs, Dr. Kotevska brings a portfolio perspective to research investment — from early-stage laboratory concepts through open-source deployment and federal policy. She is a Senior Member of IEEE, an Advisor to the IEEE USA Artificial Intelligence Policy Committee and the IEEE Computational Intelligence Society Government Activities Committee, and has responded directly to White House OSTP and NIST solicitations on privacy-enhancing technologies and AI security. She has mentored 15 PhD students across 10 universities in the United States and Europe, and 30 students and interns in total.

Leadership

  • ORNL Principal Investigator, $7M multi-institutional DOE ASCR program on privacy-preserving federated learning for scientific foundation models (2024–2027).
  • AI Safety & Security Thrust, DOE Genesis Mission — one of eight researchers on a $30M program; currently task lead for the AI safety and security taxonomy (2025–2027).
  • Lead Editor, Springer Evolving Systems special issue on verifiable and composable trust in federated and distributed learning.
  • Advisor, IEEE USA AI Policy Committee, IEEE CIS Government Activities Committee, and two AI TechX councils at the University of Tennessee.
  • Chair & Founder, IEEE Women in Engineering, East Tennessee Affinity Group (2021–present).
  • Lead organizer of four workshops and Birds of a Feather sessions at SC, TPC and CIKM; 13 organized in total.
  • 15 PhD students mentored across 10 universities in the United States and Europe; 30 students and interns in total.

Research Focus

Core research areas spanning theory and systems:

  • Federated learning — communication-efficient and privacy-preserving training across decentralized scientific data
  • Differential privacy — formal guarantees for data release and model training in sensitive domains
  • Trustworthy AI for Science — robustness, explainability, safety, and security of models under adversarial conditions

Current Projects

Genesis Mission — AI Safety & Security Thrust Active · 2025–2027
One of eight researchers on the AI Safety and Security Thrust of DOE's $30M Genesis Mission, developing threat models and safeguards for foundation models deployed across national laboratory scientific workflows. Currently task lead for the AI safety and security taxonomy. Funder: DOE ASCR.
Privacy-Preserving Federated Learning for Scientific Foundation Models PI · Active · 2024–2027
ORNL Principal Investigator on a $7M DOE ASCR program developing privacy-preserving federated training methods for scientific foundation models, with open-source implementations and cross-institutional pilots at DOE facilities. Funder: DOE ASCR, AI for Science. Project deliverables.
PETINA — Privacy Toolkit for Edge and Distributed AI Active
Open-source differential privacy library for system architectures at the edge, released through ORNL and used as the reference implementation for the ICCD 2025 practitioner guide. Repository.
PRESTO — Privacy Recommendation and Security Optimization Active · 2024–present
2025 R&D 100 Award winner. A privacy mechanism recommendation system for federated learning at scale, enabling reproducible evaluation of differential privacy, gradient protection, and secure aggregation across heterogeneous scientific data regimes. Funder: ORNL LDRD.

Get in Touch

Open to research partnerships, advisory roles, and speaking invitations.

Oak Ridge National Laboratory
Computer Science and Mathematics Division
P.O. Box 2008, Oak Ridge, TN 37831, USA
kotevskao@ornl.gov