10/27/2026 updated
100 % available
Data Science Graduate | Data Engineering & Machine Learning
Sofia, Bulgaria
Only remote
BSc Data Science (Eindhoven University of Technology)About me
I am a data science graduate that is currently studying Logic and AI in the technical university of Vienna.
Python (Programming Language)Linear RegressionMachine LearningVisualizationSQL DatabasesVisual AnalyticsConvolutional Neural NetworksXgboost
Data Engineering
Expertise in pandas, NumPy, multi-source data integration, data cleaning, missing value imputation, outlier handling, feature engineering (lagged, cyclical and spatial neighbour features), and building reproducible analysis pipelines from raw data ingestion through to final output.
Machine Learning
Proficiency with scikit-learn, XGBoost, gradient boosting, linear and penalised regression (OLS, Ridge, LASSO), Poisson regression for count data, cross validation, hyperparameter tuning, class imbalance handling, and cost sensitive learning.
Visualisation and Dashboards
Strong command of Plotly, Dash, and Matplotlib for building interactive dashboards with linked views, brushing, drill-down interfaces, and JavaScript and HTML frontends to communicate insights to non-technical stakeholders.
Programming Languages
Solid knowledge of Python, R, SQL, JavaScript, Java, C++, HTML, CSS, and LaTeX for a wide range of data science and software development tasks.
Geospatial Analysis
Experience integrating spatial boundary files, census deprivation indices, and crime records across thousands of statistical areas, and delivering interactive geospatial hotspot maps with borough-to-ward drill-down functionality.
Statistical Simulation
Ability to design simulation engines generating synthetic populations with controlled correlation structure, sparsity, and signal strength, and to run systematic variance analysis across population settings and stopping rules.
Tools and Version Control
Practical experience with Git, GitHub, GitLab, Jupyter, Google Colab, VS Code, and Excel for collaborative development and project management.
Model Interpretability
Application of SHAP values and permutation importance to audit and explain machine learning model outputs, supporting transparent and evidence-based decision making.
Expertise in pandas, NumPy, multi-source data integration, data cleaning, missing value imputation, outlier handling, feature engineering (lagged, cyclical and spatial neighbour features), and building reproducible analysis pipelines from raw data ingestion through to final output.
Machine Learning
Proficiency with scikit-learn, XGBoost, gradient boosting, linear and penalised regression (OLS, Ridge, LASSO), Poisson regression for count data, cross validation, hyperparameter tuning, class imbalance handling, and cost sensitive learning.
Visualisation and Dashboards
Strong command of Plotly, Dash, and Matplotlib for building interactive dashboards with linked views, brushing, drill-down interfaces, and JavaScript and HTML frontends to communicate insights to non-technical stakeholders.
Programming Languages
Solid knowledge of Python, R, SQL, JavaScript, Java, C++, HTML, CSS, and LaTeX for a wide range of data science and software development tasks.
Geospatial Analysis
Experience integrating spatial boundary files, census deprivation indices, and crime records across thousands of statistical areas, and delivering interactive geospatial hotspot maps with borough-to-ward drill-down functionality.
Statistical Simulation
Ability to design simulation engines generating synthetic populations with controlled correlation structure, sparsity, and signal strength, and to run systematic variance analysis across population settings and stopping rules.
Tools and Version Control
Practical experience with Git, GitHub, GitLab, Jupyter, Google Colab, VS Code, and Excel for collaborative development and project management.
Model Interpretability
Application of SHAP values and permutation importance to audit and explain machine learning model outputs, supporting transparent and evidence-based decision making.
Experience
Professional experience0 - 5 years
Freelancer since2026
Languages
BulgarianNative speakerGermanBasic knowledgeEnglishFluent
Project history
Integrated crime records, census deprivation indices and spatial boundary files from multiple open government sources across roughly 4,900 London statistical areas. Trained a Poisson objective gradient boosting model achieving a mean absolute error of 0.29, audited with SHAP and permutation importance. Delivered an interactive geospatial dashboard with borough-to-ward drill-down and documented an ethics review excluding sensitive demographic variables.
Built a Python pipeline over 246,000 Federal Railroad Administration incident records spanning 1975 to 2024, handling coordinate imputation, outlier removal, and derivation of composite risk scores. Delivered a dashboard combining a geospatial hotspot map, temporal line chart, and comparative bar chart with more than ten interdependent filters. Structured the design around Munzners nested model.
Designed a simulation engine generating synthetic populations with controlled correlation structure, sparsity, and signal strength to find the smallest sample at which error converges to within 1 percent of the population benchmark. Ran systematic variance analysis and showed penalised models stabilise with 80 to 125 fewer observations than OLS. Shipped as a dashboard with live convergence plots and tolerance bounds.