Identifying metric structures of deep latent variable models
Syrota, S., Zainchkovskyy, Y., Xi, J., Bloem-Reddy, B., & Hauberg, S. (2025). Identifying metric structures of deep latent variable models. arXiv preprint arXiv:2502.13757.
CV
Research Experience
- Since January 2025: PhD Candidate
- Technical University of Denmark, Lyngby
- Department of Applied Mathematics and Computer Science
- Supervisor: Prof. Søren Hauberg
- December 2023 – December 2024: Research Assistant
- Technical University of Denmark, Lyngby
- Department of Applied Mathematics and Computer Science
- Duties included:
- Working on uncertainty and geometry in neural networks based models using Bayesian approaches in Prof. Søren Hauberg’s lab
- Research-related software development, specifically development of matrix-free calculations based on autodifferentiation engines
- Supervisor: Prof. Søren Hauberg
Education
- M.S. in Mathematical Modelling and Computing, Technical University of Denmark, 2024
- B.S. in Mathematics-Economics, University of Copenhagen, 2020
Industry Experience
- July 2021 – March 2022: Machine Learning Engineer
- Unbox AI, Remote
- Duties included:
- Building recommender systems with SOTA Natural Language Processing models
- Downstream software development of data pipelines, microservices, and delivery of the software product
- Impact:
- Improved model accuracy by 15% through data-processing improvements and novel self-supervised training techniques
- Finalized and delivered a solution to one of the biggest grocery retail chains in Scandinavia, leading to a 10% increase in web-shop sales
- November 2021 – December 2022: Data Scientist
- PFA, Copenhagen
- Department of AI & Analytics
- Duties included:
- Development of a rule-based recommender system to improve pension and insurance offerings
- Developing a B2B-targeted product based on the recommender system to improve customer satisfaction and drive sales
- Impact:
- The recommender system won the Danish Digital Award in the Customer Experience category
- B2B side experienced 5% improvement in retention rate and 10% growth in new customers measured in contract valuations
- February 2018 – July 2020: Student Assistant
- Danish Central Bank, Copenhagen
- Department of Financial Stability (Feb 2018 - July 2020 part-time) and Department of Research (Feb 2020 - July 2020 full-time)
- Duties included:
- Development of a bank-analysis tool encompassing financial data for Denmark’s five SIFIs
- Maintenance of various SQL databases
- Research and integration of novel data sources for bank analysis
- Researching the macroeconomic effects of the COVID-19 pandemic in Denmark
- Impact:
- The bank-analysis dashboard is still in use in 2024 and shared with The Danish Financial Supervisory Authority
Skills
- Machine Learning
- Neural Networks
- Bayesian Approaches
- Natural Language Processing
- Recommender Systems
- Software Development
- Python
- SQL
- Autodifferentiation Engines
- Data Pipelines
- Microservices
- Data Science
- Statistical Analysis
- Data Integration
- Data Visualization
- Teaching
- Course Development
- Supervision
Publications
Decoder ensembling for learned latent geometries
Syrota, S., Moreno-Muñoz, P. & Hauberg, S. (2024, October). Decoder ensembling for learned latent geometries. In Geometry-grounded Representation Learning and Generative Modeling Workshop (GRaM) at ICML 2024 (pp. 277-285). PMLR.