Available for a 6-month NLP / LLM end-of-studies internship from March 2027 Contact me

Projects

From linguistic fieldwork to language engineering: from sociolinguistic surveys during my bachelor's to deep learning models in my master's, the same interest in how language is read, by humans and by machines.

Featured projects

Byzantine lead seal bearing a monogram
2025 – 2026 · Master's thesis · pair project

Reading Byzantine seal monograms

Detection + classification pipeline for Greek characters: mAP@50 0.778, 96.7% accuracy. DH2026 paper.

PyTorch · YOLOv8 · ResNet18 · ViT
Hate speech detector demo: a message classified as hateful at 87%
2026 · Personal project

Hate speech detection in French

Corpus audit (80% duplicates), weighted loss, tuned threshold: F1 0.83, precision 0.79, recall 0.88.

Transformers · CamemBERT · PyTorch · scikit-learn
Home screen of the DÉLIA app
2026 · Personal project

DÉLIA — learning NLP through play

Educational web app: 7 tracks, 143 questions, an interactive lab, Python in the browser.

JavaScript · Pyodide · HTML/CSS
Pellet reasoner inferences in Protégé
Nov. – Dec. 2025 · Semantic web

Egyptian Museum in Cairo ontology

OWL 2 + 6 SWRL rules: the reasoner dates, classifies and locates the objects.

OWL · SWRL · SPARQL · Protégé · owlready2
Semantic search engine interface
2026 · Personal project

Semantic search engine

Search by meaning: sentence embeddings and exact cosine similarity.

Sentence-Transformers · FAISS · Streamlit

Other technical projects

  • 2026OCR evaluation on 17th-century printed books — Antonomaz corpusComparing Kraken and Tesseract without a reference transcription, on 6,235 file pairs: lexicality 0.68 vs 0.62; substitution of the long s (ſ) explains much of the gap.
  • June – Sept. 2026Map of digital editing tools — CNRS / MSHS, ARIANE consortiumInternship: interactive visualisation of the project–tool network (13 projects, 38 tools) deployed on GitLab Pages, data quality control, a “tool pathways” interface. D3.js · Python · GitLab CI.

Research and critical reflection

  • Oct. – Dec. 2025Explainability, robustness and accountability of healthcare chatbotsResearch paper. Comparative study of GPT-4 and Med-PaLM 2 on 15 prompts structured along three robustness dimensions (factual, decisional, contextual); review of the regulatory framework (AI Act, GDPR, HIPAA).
  • Oct. – Nov. 2025The limits of the machineEpistemological analysis of the theoretical limits of machine learning: the No Free Lunch theorem, VC dimension, Bayes error.

Linguistic fieldwork

Bachelor's in Language Sciences, Université de Franche-Comté (now Université Marie & Louis Pasteur), Besançon.

  • Feb. – Apr. 2025Glossary of student language habitsFieldwork, transcription and sociolinguistic analysis of the vocabulary of students in Besançon; publication of a printed glossary.
  • Jan. – Apr. 2025Counter-mapping student tips in BesançonParticipatory mapping of the city from the students' point of view, especially non-native speakers: social geography, discourse and spatial practices.
  • Jan. – Mar. 2025The public debate on bullying: the Petit Nicolas caseDiscourse analysis of a media debate about bullying.
  • Jan. – Apr. 2024Linguistic variation among Malaysian students in BesançonStudy of the spoken French of Malaysian students: phonological transfer and syntactic variation linked to interference between Malay, English and French as a foreign language.