Dimosthenis Stefanidis

• Analyse user satisfaction in Azure Databricks; Deploy ML models (e.g., XGBoost) with SHAP for model interpretation, transparency and actionable insights.

• Apply quasi-experimental techniques (e.g., propensity score matching) to validate causal factors that impact user satisfaction.

• Analyse user feedback with deep learning transformer models (e.g., RoBERTa, BERTopic) to assess emotion tone and themes.

• Leverage LLMs (e.g., GPT-5.1) for summarizing and categorizing user comments.

• Data cleaning: Jargon text removal using fastText and spell checkers (99% precision); data anonymization using Microsoft Presidio to ensure user privacy (99% recall).

• Technical documentation tailored to both technical and non-technical audiences.

• Process unstructured videos (voice, text, image) using DL models for feature extraction and GNNs (PyTorch) for multimodal conversation modelling; apply XAI models (GNNExplainer, Captum); utilize GPT 4 for investment decisions and conduct moderation analysis for assessing gender biases.

• Analyse stress levels of employees using NLP, fixed effects, Difference-in-Differences, hypothesis testing, and LLMs (Mistral, Llama v2, GPT-4) on 67M tweets and 9M employee reviews; built a stress classifier using MentalRoBERTa with F1 = 0.96 and a topic classifier with F1 = 0.91 .

• Analyse disinformation and utilize LLMs (e.g. GPT-4 to simulate and identify disinformation).

• Mentored 9 data science team members.

• Implement ML solutions for forecasting energy demand (MAE = 2.20) and detecting energy demand anomalies (F1 = 0.94) using Statsmodels, VAR, Scikit-learn, and Isolation Forest.

• Process and analyse 600K photos using Face++, OpenCV, and MongoDB; conduct pattern analysis for investment decisions using advanced statistical models (Tobit, Cragg hurdle, Fixed Effects) and real-life randomized experiments with VCs, including mediation and moderation analysis.

• Design methodologies and usage scenarios, and conduct SOTA analysis for building a Big Data Platform/Marketplace for the aviation sector; develop platform components using Apache Kafka, MySQL, Docker CI/CD, and Java Spring Boot.

• Develop distributed and multithreaded crawlers using Selenium, Docker, Kafka, and MongoDB.