Last updated May 2026. For a PDF version, click here.

Education

• Ph.D in Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA, USA, 2025
• M.Sc. in Applied Mathematics, Faculty of Sciences, University of Novi Sad, Novi Sad, Serbia, 2018
• B.Sc. in Applied Mathematics, Faculty of Sciences, University of Novi Sad, Novi Sad, Serbia, 2016

Experience

• Sep 2025 - present: Post-doctoral researcher at École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland
  • Advisor: Prof. Ali H. Sayed
• Aug 2020 - Jun 2025: Graduate research assistant at Carnegie Mellon University, Pittsburgh, PA, USA
  • Advisor: Prof. Soummya Kar
• Sep 2019 - Dec 2019: Visiting researcher at University of Pennsylvania, Philadelphia, PA, USA
  • Advisor: Prof. Alejandro Ribeiro
• Nov 2018 - Jun 2020: Junior researcher at Faculty of Sciences, University of Novi Sad, Novi Sad, Serbia
  • Advisor: Prof. Dušan Jakovetić

• July 2018 - Oct 2018: Junior researcher at BioSense Institute, University of Novi Sad, Novi Sad, Serbia

• Feb 2018 - July 2018: Visiting researcher at Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia
  • Advisor: Prof. Blaž Zupan

Publications (in chronological order, * = equal contribution)

High-Probability Convergence in Decentralized Stochastic Optimization with Gradient Tracking

Armacki, A., Cai, H., & Sayed, A. H. (2025). High-Probability Convergence in Decentralized Stochastic Optimization with Gradient Tracking. In arXiv:2605.00281

Tight Long-Term Tail Decay of (Clipped) SGD in Non-Convex Optimization

Armacki, A., Bajović, D., Jakovetić, D., Kar, S., & Sayed, A. H. (2025). Tight Long-Term Tail Decay of (Clipped) SGD in Non-Convex Optimization. In arXiv:2602.05657

High-probability Convergence Guarantees of Decentralized SGD

Armacki, A., & Sayed, A. H. (2025). High-probability Convergence Guarantees of Decentralized SGD. In arXiv:2510.06141

Optimal High-probability Convergence of Nonlinear SGD under Heavy-tailed Noise via Symmetrization

Armacki, A., Bajović, D., Jakovetić, D., & Kar, S. (2025). Optimal High-probability Convergence of Nonlinear SGD under Heavy-tailed Noise via Symmetrization. In arXiv:2507.09093

Toward Understanding the Improved Robustness to Initialization in Distributed Clustering

Armacki, A., & Kar, S. (2025). Toward Understanding the Improved Robustness to Initialization in Distributed Clustering. In 33rd European Signal Processing Conference.

High-probability Convergence Bounds for Online Nonlinear Stochastic Gradient Descent under Heavy-tailed Noise

Armacki, A., Yu, S., Sharma, P., Joshi, G., Bajović, D., Jakovetić, D., & Kar, S. (2025). High-probability Convergence Bounds for Online Nonlinear Stochastic Gradient Descent under Heavy-tailed Noise. In 28th International Conference on Artificial Intelligence and Statistics.

A Unified Framework for Gradient-based Clustering of Distributed Data

Armacki, A., Bajović, D., Jakovetić, D., & Kar, S. (2025). Distributed Center-Based Clustering: A Unified Framework. In IEEE Transactions on Signal Processing, vol. 73, pp. 903-918, doi: 10.1109/TSP.2025.3531292.

Distributed Gradient Clustering: Convergence and the Effect of Initialization

Armacki, A.*, Sharma, H.*, Bajović, D., Jakovetić, D., Chakraborty, M., & Kar, S. (2024). Distributed Gradient Clustering: Convergence and the Effect of Initialization. In 58th Asilomar Conference on Signals, Systems, and Computers.

Large Deviations and Improved Mean-squared Error Rates of Nonlinear SGD: Heavy-tailed Noise and Power of Symmetry

Armacki, A.*, Yu, S.*, Bajović, D., Jakovetić, D., & Kar, S. (2024). Large Deviations and Improved Mean-squared Error Rates of Nonlinear SGD: Heavy-tailed Noise and Power of Symmetry. In arXiv:2410.15637

A One-shot Framework for Distributed Clustered Learning in Heterogeneous Environments

Armacki, A., Bajović, D., Jakovetić, D., & Kar, S. (2024). A One-Shot Framework for Distributed Clustered Learning in Heterogeneous Environments. In IEEE Transactions on Signal Processing, vol. 72, pp. 636-651, doi: 10.1109/TSP.2023.3343561.

Communication Efficient Model-aware Federated Learning for Visual Crowd Counting and Density Estimation in Smart Cities

Armacki, A., Milošević, N., Bajović, D., Kar, S., Jakovetić, D., Bakhtiarnia, A., Esterle, L., Muscat, A., & Festi, T. (2023). Communication Efficient Model-aware Federated Learning for Visual Crowd Counting and Density Estimation in Smart Cities. In 31st European Signal Processing Conference, Helsinki, Finland.

Personalized Federated Learning via Convex Clustering

Armacki, A., Bajović, D., Jakovetić, D., & Kar, S. (2022). Personalized Federated Learning via Convex Clustering. In IEEE International Smart Cities Conference (ISC2), Pafos, Cyprus, pp. 1-7, doi: 10.1109/ISC255366.2022.9921863.

Gradient Based Clustering

Armacki, A., Bajović, D., Jakovetić, D., & Kar, S. (2022). Gradient Based Clustering. In proceedings of the 39th International Conference on Machine Learning, PMLR, 162:929-947.

Distributed Trust-region Method with First Order Models

Armacki, A., Jakovetić, D., Krejić, N., & Jerinkić, N. K. (2019). Distributed Trust-Region Method With First Order Models. In IEEE EUROCON 18th International Conference on Smart Technologies, Novi Sad, Serbia, pp. 1-6, doi: 10.1109/EUROCON.2019.8861739.

Awards

• IBM PhD Fellowship - Finalist. One of three nominees by the Electrical and Computer Engineering Department, Carnegie Mellon University, 2023
• Dean’s Fellowship - Electrical and Computer Engineering Department, Carnegie Mellon University, 2020
• Annual Award of the Mathematical Institute of the Serbian Academy of Sciences and Arts for M.Sc. Students in the Field of Computing - Finalist, 2019

Academic Service

• Conference reviewer: IEEE ISIT, AISTATS, ICML, NeurIPS.

• Journal reviewer: IEEE Transactions on Signal Processing, Journal of Machine Learning Research.

• Student supervision:

  1. Haoyuan Cai (PhD student, EPFL, fall 2025-present): actively supervising and collaborating since joining prof. Sayed’s group. The supervision resulted in publication.

  2. Himkant Sharma (research intern, Carnegie Mellon University, summer 2024): supervised for the duration of internship in prof. Kar’s group. The supervision resulted in publication.

• Teaching: Linear Systems, Fall 2022 and 2023.

• Student council for faculty hiring, Electrical and Computer Enigneering Deparment, Carnegie Mellon University, 2023.