Dorina Ababii

Resume in WET format with GitHub Pages

I developed an interactive, accessibility-compliant resume in the Web Experience Toolkit (WET) format, hosted on GitHub Pages. Utilizing HTML, CSS, and JavaScript, I created a responsive design. Through this project, I enhanced my web development skills, gained proficiency in version control with GitHub, and successfully deployed websites on GitHub Pages.
Tools: GitHub Pages, HTML, CSS

October 2024

AI-Driven Compliance Automation

Project at University of Zürich, Compliance Department. Developed an AI-driven solution Proof-of-Concept to streamline, harmonize, translate, and update legal client forms, leveraging Large Language Models (LLMs), Optical Character Recognition (OCR), and Natural Language Processing (NLP). The project enhanced workflow automation for risk and compliance, delivered in collaboration with a cross-functional team.
Tools and Techniques: Applied AI, LLMs, OCR, NLP, and data analytics tools to optimize compliance operations.

September 2024

Master Thesis: Machine Learning in Asset Pricing

Master thesis conducted within the University of Zürich of Applied Sciences. The objective was to evaluate the application of OLS, LSTM, and CatBoost models for asset pricing in both regional and global portfolios. The project involved analyzing international stock data (2000–2021), segmented into development and testing phases, using stock-specific data from Bloomberg and Refinitiv, alongside macroeconomic indicators from FRED. Results demonstrated that LSTM models significantly outperform traditional linear models in predictive accuracy, showcasing explainable AI’s potential in enhancing portfolio strategies.
Tools and Techniques: LSTM, CatBoost, OLS, Python, Bloomberg, Refinitiv, FRED
Challenges: Managed complexities in model selection, data scaling, and feature selection.

June 2021 - December 2021

Text Sentiment Classification and Analysis

Project conducted at the University of Zürich of Applied Sciences focused on enhancing sentiment analysis through Natural Language Processing (NLP). Developed and tested Multinomial Naive Bayes and neural network models to classify sentiment using a Kaggle dataset, with potentiL practical applications in marketing, finance, and social media decision-making.
Tools and Techniques: Python, NLP methodologies, Multinomial Naive Bayes, neural networks, model evaluation metrics
Key Learnings: Strengthened expertise in sentiment analysis and text classification, applying machine learning for impactful, real-world insights.

June 2021 - September 2021

Finance and Interpretability of Machine Learning Models

Project completed at the Zurich University of Applied Sciences under the guidance of Dr. Bledar Fazlija. This research focused on developing interpretable machine learning models tailored for credit card fraud detection in financial services. Using Logistic Regression, Decision Trees, and CNNs, the project tackled the challenge of balancing predictive accuracy with interpretability, essential for industry compliance and trust. Techniques employed included feature importance, partial dependence plots, and LIME, offering a transparent approach to meet regulatory and ethical standards.
Tools and Techniques: Logistic Regression, Decision Trees, feature importance, partial dependence plots, LIME

February 2021 - June 2021

Financial Modeling Project: Robust Statistical Analysis

Project conducted at the University of Zürich of Applied Sciences focused on developing a reliable framework for financial data analysis through advanced statistical techniques. The objective was to create a comprehensive database of stock and economic data and build models to analyze stock returns using optimized OLS regression. Ensured model robustness by testing for heteroskedasticity, multicollinearity, and autocorrelation, enhancing data integrity and predictive accuracy.
This project honed my skills in financial data diagnostics, model optimization, and statistical reliability assessment.
Challenges: Achieving model stability, accuracy, and diagnostic rigor in complex financial datasets.
Tools: Python

February 2021 - June 2021

Semester Project: Switzerland’s Financial System and Real Economy Interplay during the Pandemic

Research project at the University of Zurich of Applied Sciences focused on analyzing Switzerland’s economic response to financial disruptions induced by the pandemic. This study aimed to develop and test financial models that forecasted business cycles and assessed the pandemic’s impact on Switzerland’s real economy. The project highlighted advanced insights into crisis-driven market changes through macroeconomic analysis and financial forecasting.
Challenges: Advanced data modeling, macroeconomic trend forecasting, crisis impact assessment.

October 2020 - December 2020

CFA Institute Research Challenge

Academic project undertaken at the University of Zürich of Applied Sciences, focused on analyzing a single stock to assess its buy, hold, or sell potential. The objective was to apply financial theories and valuation techniques to evaluate stock performance, risk profile, and growth prospects. This experience provided an in-depth understanding of financial modeling, risk assessment, and data-driven investment recommendations within a competitive setting.
Skills Applied : Financial modeling, Excel, risk analysis techniques, stock valuation, risk profiling, and investment recommendations.
Insights : Strengthened expertise in financial theories and investment decision-making through rigorous data analysis.

October 2020 - November 2020