At Profile Software, I evaluated AI-generated Greek speech-to-text output for accuracy, fluency, and grammatical correctness. I identified recurring morphology and named-entity errors and performed linguistic quality checks on datasets used to train and validate ASR systems.
I built a sentiment analysis project that fine-tuned and compared BERT-base and DistilBERT on the IMDb 50K dataset. BERT-base achieved 93.61% accuracy and 93.66% F1, while DistilBERT achieved 93.10% accuracy and 93.09% F1.
I also designed and implemented a task-oriented conversational AI system using Rasa Open Source. It includes intent classification, entity extraction, slot filling, dialog policies, REST API integrations, and fallback handling.
My MSc in Language Technology (NLP) at the National & Kapodistrian University of Athens focuses on NLP, computational linguistics, machine learning foundations, transformers, and speech and text processing. My background in Greek language and literature informs how I approach linguistic precision in applied machine learning.
