I'm a master's student in Language Technology at the University of Gothenburg, with a BEng in Electronic Engineering from Greece. I work on spoken dialogue systems, NLP and ML. I'm interested in where software meets real-world hardware.
Before the master's I worked with hardware. During a six-month internship in the maintenance department at Nexans, a cable factory, I repaired motors, sensors and power-supply boards alongside the maintenance team. Later, as a telecom technician, I connected homes to the internet over fibre, DSL and 4G/5G.
- Elisa: a spoken movie recommender. You tell it who is watching and what you're in the mood for, and it suggests films from about 23,000 titles until one fits.
- CODer: can you tell code written by a person from code written by an LLM? Fine-tuned code encoders reach up to 0.95 macro-F1. The main finding is that code an LLM has edited is much harder to detect than code it wrote from scratch.
- Multilingual tokenization: which tokenizer suits English, Turkish and Chinese best? One small language model per tokenizer, compared by bits per character.
Looking for a master's thesis from January 2027, and a first engineering role in the Nordics after I graduate.