Aurélien Callens

🌍 R&D Data Scientist | Expert in Environmental & Geospatial Data I help companies, startups, and research labs leverage environmental and geospatial data for applications in meteorology, precision agriculture, climate risk management, and ecology. 📌 Areas of Expertise - Geospatial data acquisition & processing (sensor data or remote sensing) - Spatiotemporal analysis & modeling - Data quality, validation & reliability - AI & Machine Learning applied to environmental sciences - Deployment & industrialization of data solutions 🛠 Tech Stack 💡 Python, Pytorch, TensorFlow, Xarray, Kedro, MLflow, Spark, PostgreSQL 📍 Available for freelance projects – Remote & hybrid in Toulouse or Bordeaux 🚀

As the first R&D Data Scientist in this AgTech scale-up providing connected weather stations to farmers, I defined and led the data and meteorology R&D roadmap. My goal: fully leverage collected weather data, combining company strengths and scientific advances to deliver value to farmers. Results & Impact → Designed and deployed an anomaly detection algorithm to monitor, in real time, the quality of meteorological data from 30,000 weather stations. → Developed a deep learning algorithm merging precipitation data from multiple sources (stations, radar, satellite) to enhance coverage across Europe (100% vs. 25-75% from individual networks). Achievements 🔹 Defined scientific objectives, conducted literature reviews, developed methodologies, validated results, and ensured knowledge dissemination. 🔹 Engineered and implemented advanced machine learning models in Python. 🔹 Deployed training and inference pipelines for large-scale data processing. 🔹 Collaborated closely with Data Engineering teams to industrialize solutions.

Leading a Research Project on Coastal Depth Prediction from Video Surveillance Results & Impact → Deep learning model achieving a 45 cm RMSE, matching state-of-the-art performance with promising improvements for coastal monitoring. →Strengthened collaboration with a strategic partner (IRT Toulouse) through technical follow-ups and presentations. Achievements 🔹 Defined research objectives and project scope in alignment with partner needs. 🔹 Designed a data pipeline to collect, clean, and integrate heterogeneous sources (weather stations, buoys, video surveillance). 🔹 Developed and trained the deep learning model using Keras, including hyperparameter optimization. 🔹 Conducted activation and uncertainty analysis to improve model explainability. 🔹 Built an interactive dashboard (Dash) for monitoring key performance metrics.

As part of my PhD thesis, "Statistical Learning for Coastal Risks Assessment", I led a research project aimed at developing innovative data-driven methodologies to improve coastal risk assessment. Results & Impact → 30% reduction in prediction errors of an operational wave model through data assimilation. → Automated collection of storm impact data using a computer vision model, achieving 87% accuracy. → Operational deployment of a Bayesian network for real-time coastal flood prediction, enhancing decision-making and early warning systems. → Three scientific publications in international journals and presentations at three national/international conferences. Achievements 🔹 Defined scientific objectives, conducted literature reviews, developed methodologies, validated results, and ensured scientific dissemination. 🔹 Designed and implemented multiple data pipelines to collect, clean, and integrate heterogeneous data sources (weather stations, buoys, video surveillance). 🔹 Analyzed the seasonal variability of waves using 30 years of spatiotemporal data from the Atlantic Ocean.

Development of a robust linear regression method to analyze time series exhibiting heteroscedasticity and outliers. Results & Impact → Academic paper : Callens, A., Wang, Y., Fu, L. et al. Robust Estimation Procedure for Autoregressive Models with Heterogeneity. Environ Model Assess (2020). https://doi.org/10.1007/s10666-020-09730-w Achievements 🔹 Defined scientific objectives, conducted literature reviews, developed the methodology, validated results, and ensured scientific dissemination. 🔹 Implemented the method in R and distributed the code via an R package on the Comprehensive R Archive Network (CRAN).

Analysis of physical and climatic forcings impacting the transport of micropollutants in the Adour estuary. Results & Impact → Identification of key physical and climatic forcings influencing the transport of micropollutants in the Adour Estuary through advanced statistical analyses. → Development of an interactive tool to visualize and explore the results of time series and extreme data analyses. → Contribution to coastal storm modeling research through a collaborative study published on the use of copulas to characterize extreme wave events (http://www.jcronline.org/doi/abs/10.2112/SI85-161.1). Achievements 🔹 Statistical analysis of time series and extreme data related to micropollutant transport. 🔹 Modeling of storm-induced wave characteristics using copulas in R.

Species identification to train a deep learning algorithm

Investigation on the relationship between the nematode S.carpocapsae and several strain of bacteria (Xenorhabdus genus).

Full time contract during holliday and part time contract (week-end) during the academic year.