PhD engineer with 7+ years of experience in Python, machine learning, data science and scientific software development. My background is in scientific R&D, where I worked on complex experimental data, machine learning pipelines and user-facing software. Over time, my work has expanded toward backend development, APIs, deployment, automation and more recently generative AI. I enjoy building tools that are useful, maintainable and understandable by the people who use them.

Python • FastAPI • Flask • REST APIs • SQL • Docker • Linux • Git / CI • pytest

scikit-learn • PyTorch • TensorFlow / Keras • Pandas • NumPy
Machine learning • Deep learning • RAG • Embeddings • LLMs • Ollama • Whisper • Model evaluation

Scientific and high-dimensional data • HDF5 • Data visualization
Workflow automation • n8n • Monitoring • Technical documentation • Requirements analysis

Python • FastAPI • scikit-learn • pytest • Docker

A reproducible ML API for molecular classification from SMILES, including training, validation, testing and documentation.

Python • FastAPI • Ollama • Embeddings • Hybrid Search

A local RAG system for scientific literature with source-grounded answers, citations, hybrid retrieval and comparison of retrieval strategies.

Python • Whisper • LLMs

A local application for meeting recording, speech-to-text transcription and structured note generation.

🔒 Note: Most of my recent industrial R&D work is confidential. The examples below illustrate some of the technical problems I have worked on across software, machine learning and scientific data.

• Machine learning for noisy high-resolution signals
  Designed preprocessing, data augmentation and model evaluation workflows for PFAS identification from complex ¹⁹F NMR spectra, including simulated spectra and noise injection. Achieved 96% precision and 95% recall, with a strong focus on error analysis and robustness.
• Python SaaS development for expert users
  Developed and maintained Python/Flask applications within a SaaS platform, translating expert user needs into specifications and maintainable software features. Contributed to Docker/Linux deployment, Git/CI workflows, monitoring, documentation and technical support.
• Large-scale data processing and memory-efficient workflows
  Worked on 2D FT-ICR MS deconvolution for very large datasets, addressing RAM limitations through out-of-core processing with HDF5, block-wise data loading and dynamic resolution selection.
• Interactive tools and data visualization
  Built interactive scientific dashboards and analysis tools with Bokeh, including volcano plots and spectral exploration tools, with an emphasis on making complex outputs understandable to non-programming users.
• Deep learning for automated signal attribution
  Designed and trained neural networks for automated NMR signal attribution, reaching 89.5% accuracy on experimental data and outperforming the previous benchmark tool.
• Reproducible ML and data workflows
  Built reproducible pipelines for data preparation, training, hyperparameter optimization, validation and model evaluation, with attention to traceability, documentation and FAIR data practices.
• Collaborative R&D and technical project work
  Contributed to multidisciplinary R&D projects, technical documentation, funding proposals and coordination between scientific, technical and user-facing stakeholders.

My background is rooted in scientific R&D — including NMR, FTICR-MS, molecular data and biotechnology — but the core of my work has increasingly focused on Python development, machine learning, data-intensive software and user-oriented tools. I am particularly interested in applying these skills to broader technical and industrial problems.

Python development • Backend / APIs • Machine Learning • Applied AI • RAG / LLM applications • Data-intensive software • Scientific software

• LinkedIn: www.linkedin.com/in/laura-duciel
• Email: laura.duciel@hotmail.fr
• Location: Illkirch-Graffenstaden, France

"Transforming complex data into actionable ML models."