My CV

Data Engineer

Machine Learning Engineer with a Master's degree in Data & AI and signal processing. I design AI solutions from prototyping to validation, with a particular interest in embedded and connected systems.

Experience

  1. R&D Intern — ML for musical improvisation

    SCRIME · Bordeaux

    • Developed LSTM and Transformer generative models for real-time audio generation: inference latency, runtime stability and output quality.
    • Trained and evaluated models on large-scale symbolic music datasets.
    • Designed and optimized real-time AI pipelines for interactive applications (low latency, runtime reliability).
    • Studied model deployment constraints in resource-constrained interactive environments.
  2. R&D Intern — Computer vision

    Alten · Sèvres

    • Developed an end-to-end solution for pedestrian intention detection in urban environments.
    • Implemented, evaluated and deployed CNN and Transformer models (PyTorch, TensorFlow, OpenCV) within complete machine learning pipelines.
    • Optimized computational efficiency for production-oriented deployment.
  3. NLP / data processing intern

    France Télévisions · Paris

    • Built training datasets for translation models.
    • Developed ingestion and preprocessing pipelines for NLP data and relational databases.

Education

  1. Master's in Intelligent Systems Engineering

    Sorbonne Université · Paris

    AI, machine learning applied to audio and vision, deep learning, reinforcement learning, NLP.

  2. Master's in Robotics, Linear Control and AI

    Sorbonne Université · Paris

  3. Bachelor's in Electronics, Electrical Energy and Automation

    Sorbonne Université · Paris

Skills

Languages
Python, SQL, C++, C
Generative AI & RAG
LLM, RAG, LangChain, Chroma (vector database), Hugging Face embeddings, retrieval evaluation (Precision, Recall, MRR, Hit Rate)
Machine learning
PyTorch, TensorFlow, scikit-learn, Transformers, ONNX, OpenCV
Development & industrialization
FastAPI, REST APIs, Git/GitHub, Docker, Linux, testing (pytest)
Audio & signal
Librosa, ffmpeg, JUCE, MIDI

Projects

RAG system for football knowledge retrieval

Local RAG pipeline (LangChain, Chroma, Hugging Face embeddings, local LLM): ingestion and chunking of Wikipedia articles (FIFA, World Cup), similarity search, retrieval evaluation with Precision, Recall, MRR and Hit Rate.

VocalTwin — Articulatory speech analysis and video validation

Web application (FastAPI) that estimates articulatory landmarks from an audio file (SPARC model) and animates them in sync with the sound. Audio extraction with ffmpeg, caching analyses by file hash, profile-video validation mode. Automated tests (pytest), architecture, data-schema and limitations documentation.

Audio deepfake detection

Fine-tuning and deployment of a HuBERT Transformer model for authentic vs. synthetic speech classification.

Languages