Expériences professionnelles
Data science
SWISSCOM AG - CDD
De Février 2023 à Août 2023
● Researched and implemented state-of-the-art explainability methods for creating explanations of black-box models.
● Developed a new explainability method using the Attention Mechanism to explain deep learning models trained with
tabular data—a full pipeline implemented in PyTorch.
● Participated in different activities to promote knowledge sharing across team members and shared the work with other
colleagues in the company through different presentations every week.
Student researcher
LHTC lab EPFL , Lausanne - CDD
De Octobre 2022 à Février 2023
● Worked on two projects for creating a non-invasive method for acquiring blood pressure parameters from time-series
data representing brachial waveforms.
● Implemented CNN models that predict the aortic systolic blood pressure (aSBP) and the cardiac output (CO) from the
brachial pressure waveforms.
● Developed end-to-end pipeline in PyTorch, comprising pre-processing of time series and training/testing CNN models.
● Created a non-invasive method showing promising results with an R-squared result of 0.92.
● Tested on synthetic data and performed adjustments to adapt it to real data in an environment with reduced data samples.
Data scientist
DOMOSAFETY SA - CDD
De Février 2022 à Août 2022
● Created a dataset from time-series data using Pandas and TFRecord to be used for analytics and machine learning tasks.
● Developed prototypes (i.e., VICReg) using PyTorch and packaged the code to integrate into the company’s code stack.
● Developed visualisation methods (UMAP, PCA) using JavaScript and Plotly to visualise and evaluate the results from
the machine learning models.
● Implemented a self-supervised model that creates embeddings of time-series data from physiological measurements,
enabling anomaly detection of artifacts.
Parcours officiels
Langues
Anglais - Technique
Français - Technique
Macédonien du nord - Langue maternelle
Compétences
Centres d'intérêt
- swimming
- hiking
- winter sports