AI Engineer specialising in agentic, GenAI, NLP and computer vision. I design the models as much as the data pipelines and the deployment chain that carry them all the way to users.
I'm an AI Engineer based in the Paris region. I design and ship machine learning and generative AI systems, along with the data pipelines that feed them.
Since January 2024, at EstimerMonCommerce.fr, I shipped a production OCR pipeline, led R&D on open-source LLMs, and built the ingestion and geospatial analysis chains the product runs on. On 5 October 2026 I join Innovorder on a permanent contract as AI Solution Engineer.
On the academic side: a Master's in Computer Science — Applied AI & Data at Epitech, preceded by a Bachelor's in systems, networking and IT infrastructure. That double grounding explains what interests me most — everything that happens between a model that works in a notebook and a service that holds up in production.
In 2025 I was a MasterDevFrance finalist in the AI and algorithm hackathons.
Innovorder
I'm joining Innovorder on a permanent contract to design and integrate production AI solutions: autonomous agents, process automation, and connection to existing business tools.
EstimerMonCommerce.fr · Full-remote
Three years owning the AI and data side of a SaaS valuation product, from research prototype through to production service.
LTG Services
Designed, built and shipped a cross-platform iOS/Android professional networking app, owning architecture and UX through to store deployment.
Epitech Technology · Nancy
A programme centred on end-to-end development and deployment of AI systems for real-world applications.
Epitech Technology · Nancy
An engineering programme emphasising systems, networking and practical IT infrastructure.
The ground I'm most comfortable on, and what I actually do there.
Prototyping and experimentation: evaluation protocols, comparative model analysis, rigorous measurement of results.
Training, transfer learning and fine-tuning, evaluation metrics, handling class imbalance and overfitting.
CNNs for vision, Transformers and BERT-based models for language, deep architectures applied to real data.
Ingestion and scraping pipelines, cleaning and normalisation, scalable storage and modelling-ready datasets.
Getting models to production: orchestration, observability with LangFuse, monitoring and deployment pipelines.
Deployment and execution on cloud platforms, serverless orchestration, GPU-accelerated inference.
Python for AI and data, full-stack React and PHP, Bash for automation.
Framing business needs, performance/cost trade-offs, iterative delivery in close contact with the product.
A few pieces of applied research, then my activity over the last twelve months and my public repositories.
A comparative analysis of CNN architectures (EfficientNet vs. ResNet) using transfer learning to classify skin pathologies from clinical imagery. Engineered a robust preprocessing pipeline — stain normalisation, advanced data augmentation — designed to mitigate dataset imbalance and overfitting.
Explored and implemented state-of-the-art NLP architectures, focusing on Transformers and BERT-based models for Named Entity Recognition. Ran transfer learning and fine-tuning experiments to adapt pre-trained models to domain-specific datasets, evaluated against rigorous NLP metrics, with deep work on attention mechanisms and sequence labeling.
An automated ingestion pipeline combining spectral analysis, audio normalisation and waveform preprocessing to optimise inputs for neural inference. A scalable, GPU-accelerated inference system orchestrated serverless, sized to handle high-dimensional audio data at low latency.
My GitHub and GitLab activity over the last twelve months, and my pinned repositories.
End-to-end NLP pipeline that converts French natural language queries into optimal train itineraries using CamemBERT (intent classification + NER) and A* pathfinding over the SNCF GTFS network.
Open-source AI audio stem separation SaaS powered by BS-Roformer and serverless GPU.
A convolutional neural network trained on the HAM10000 dataset for multi-class skin lesion classification.
Automatic detection of pneumonia from X-ray images using Machine Learning & Deep learning techniques.
Docs search system w/ hybrid RAG (and LLM)
Platform collecting crypto prices via Binance, processing streams with Kafka/Spark, storing in TimescaleDB, and visualising in real time via React.
I'm happy to talk about AI in production: agents, applied GenAI, data pipelines and MLOps.
A question, a project, or simply the urge to dig into a technical topic — write to me, I answer.