Education

Technical University of Munich, Master’s Degree in Mathematics in Data Science (10/2021 – 07/2024)
- Awarded the DAAD Master Scholarship for academic performance and research potential.
Igor Sikorsky Kyiv Polytechnic Institute, Bachelor of Science in System Analysis (09/2017 – 06/2021)
- Received a merit-based scholarship for academic excellence, ranking in the top 15% of students.

Publications and Preprints

Amortized Structured Stochastic Variational Inference for Gaussian Process Latent Variable Models (2026)
- Submitted to ICML 2026
- Authors: Maksym Tretiakov, Sarah Filippi, Vincent Fortuin, Ruth Misener, Ruby Sedgwick, James Odgers
PAC Bayes Bounds Evaluation Framework (2025)
- PDF: github.com/fortuinlab/pacbb/blob/main/doc/pacbb.pdf
- Authors: Yauheni Mardan, Maksym Tretiakov, Alexander Immer, Vincent Fortuin
A new atlas to study embryonic cell types in Xenopus (2024)
- Developmental Biology Journal
- Authors: Kseniya Petrova, Maksym Tretiakov, Aleksandr Kotov, Anne H. Monsoro-Burq, Leonid Peshkin

LMU Munich (Munich Uncertainty Quantification AI Lab), Ph.D. Candidate in Statistics (03/2026 – Current)
- Supervisor: Professor David Rugamer
- Lab website: https://www.muniq.ai/
Scalable Capital, Quantitative Developer (01/2024 – Current)
Helmholtz.AI (ELPIS Lab), Research Project (04/2025 – Current)
- Supervisors: Dr. James Odgers, Professor Vincent Fortuin
- Developed a Structured Stochastic Variational Inference method for GP-LVMs, targeting limitations of mean-field variational approximations in uncertainty estimation.
- Derived a structured variational objective and evaluated uncertainty estimates against mean-field baselines in an end-to-end PyTorch framework.
Harvard Medical School (Kirschner Lab), Research Project (10/2022 – 07/2024)
- Supervisor: Dr. Leon Peshkin
- Implemented machine learning methods for single-cell transcriptomics, including clustering and classification algorithms for cell type annotation.
- Built scalable pipelines for unbiased cell type assignment and validation.
Institute of Mathematics of National Academy of Sciences of Ukraine, Research Project (11/2020 – 05/2021)
- Supervisor: Professor Andrey Pilipenko
- Developed and compared Hidden Markov Models and deep learning methods (RNN, LSTM) for automatic chord recognition in music.
- Implemented audio preprocessing pipelines (Fourier, HPS, CQT transforms) and evaluated models on real-world song datasets.