M2 student in AI, Machine Learning & Data Science at Sorbonne Université (Master MIND)
NLP · LLMs · Machine Learning · Data
Looking for a 6-month internship in AI / Data Science / Machine Learning, starting March 2027. Open to research labs and companies, anywhere in France (on-site, hybrid or remote).
- I'm in my second year of the Master MIND (formerly DAC) at Sorbonne Université, which covers artificial intelligence, machine learning and data science.
- During my research internship at LIP6, I studied when LLM constrained decoding engines violate JSON Schemas, and built models that predict these errors on unseen schemas.
- What I enjoy most is turning a messy real-world question into a sound experimental setup: building the dataset, choosing a validation that doesn't leak, comparing models honestly, and explaining the results.
- Currently learning more about LLMs, deep generative models and explainable AI.
- Languages: French (C2), English (B2), Arabic (native).
| Project | What it's about | Highlights |
|---|---|---|
| Constrained decoding error prediction Research internship · LIP6 |
Predicting when Guidance, Outlines and XGrammar accept invalid or reject valid JSON on JSONSchemaBench | ~76k test decisions, leakage-free grouped splits, LightGBM / RF, interpretable rules · ROC-AUC up to 0.99 |
| BirdCLEF+ 2026 Kaggle competition |
Multi-label species detection in audio soundscapes (234 classes) | Handcrafted audio features + LightGBM, co-occurrence correction, domain adaptation · public macro-AUC 0.783 |
| Chess Elo prediction & player profiling | Estimating a player's Elo from a few Lichess games and describing their style | 56k games, ~215 behavioural features, few-shot setting, DeepSets embeddings · MAE ≈ 100 Elo |
| NLP: sentiment & authorship | Movie-review sentiment and Chirac vs Mitterrand speaker attribution | RoBERTa head+tail + SVM ensemble, CamemBERT with document context · F1 93.1 |
| ChatPhone: RAG chatbot | Recommending smartphones grounded in Amazon reviews | Sentence-BERT + FAISS retrieval, sentiment-filtered indexes, Qwen3 via Ollama, FastAPI / Flask |
| Passage retrieval reproduction | Reproducing Retrieving Passages and Finding Answers on TREC-DL 2019 | QL, SDM, positional models, relevance feedback, character-level evaluation (PyTerrier) |
| Forest fire prediction in Béjaïa Bachelor's final project · LSI lab, USTHB |
Predicting fire risk from vegetation, soil, weather and topography | 1.6M-row geospatial dataset, extreme imbalance, XGBoost, GIS risk map, Django / Leaflet app |
More projects (optimisation, databases, mobile)
| Project | Highlights |
|---|---|
| Robot pathfinding: BFS + MILP | State-space graph and optimal BFS, complexity study, instance generation with an integer linear program (Gurobi), Flask app |
| Big Match: SQL recommendation | PostgreSQL schema, 200k rows of synthetic data, window functions, profile and event recommendation written in SQL |
| PLV e-commerce app Freelance 2024 |
Flutter app: authentication, profile and 4-step checkout with a REST API, GetX and Firebase (11 merged PRs) |
Languages
Machine learning & deep learning
NLP, LLMs & information retrieval
Data, optimisation & tools
| When | What |
|---|---|
| 2026 – 2027 | M2 MIND: AI, Machine Learning & Data Science, Sorbonne Université |
| Jun – Jul 2026 | Research intern, LIP6 (Sorbonne Université): error analysis and prediction for LLM constrained decoding |
| 2025 – 2026 | M1 MIND, Sorbonne Université |
| 2024 – 2025 | Bachelor's degree (L3) in Computer Science, Université Paris Cité |
| Jun – Aug 2024 | Freelance mobile developer: Flutter e-commerce app |
| 2024 | Bachelor's degree in Software & Information Systems Engineering, USTHB, Algiers |
Relevant coursework
M2 (2026 – 2027): Large Language Models · Deep Learning · Advanced Deep Learning: Generative Models & AI for Science · eXplainable AI · Pattern Recognition for Image Analysis · Large-Scale Data Analytics · Data & Knowledge Graphs · Methodology in Data Science & Research
M1 (2025 – 2026): Machine Learning · Information Retrieval & NLP · Data Science, Learning & Applications · Modelling, Optimisation, Graphs & Linear Programming · Research project (PMIND)
I'm happy to chat about internships, NLP/LLM projects or data science in general. Reach me at wafaaberrais2@gmail.com or on LinkedIn.