Abdelbari Bouzarkouna
Lead Data Scientist
Salary / Daily rate
Chatou, France
Freelance, open to permanent contract
Can come onsite
Skills
Languages
About me
Intro
Lead Data Scientist with extensive experience in industrial AI and ML solutions, specializing in yield forecasting and manufacturing optimization. Successfully led development of data-driven models managing €1B in commodities across 200,000 hectares at Tereos. Expert in end-to-end ML deployment, from cloud architecture to production, combining statistical modeling with business impact in Agriculture and Energy sectors.
Work experience
Tereos
Jan 2020 - Present
6 years 10 months
Paris, France
Lead Data Scientist
Apr 2023 - Present
3 years 7 months
Paris, France
At Tereos, I led the development of the company's first data-driven yield forecasting model, covering 200,000 hectares and managing a commodity valued at €1 billion. Key achievements: Needs assessment: Collaborated with the agricultural and IT departments to understand project challenges and operational requirements. Data collection and processing: Gathered and processed diverse datasets, including agronomic, climatic, and historical data, to build a robust modeling foundation. Modeling: Designed and fine-tuned a predictive model to forecast yields, incorporating advanced algorithms tailored to the agricultural sector. Satellite imagery integration: Utilized satellite data to estimate field coverage rates, assess foliar disease impacts, and monitor harvest progress. Results delivery: Effectively communicated insights and recommendations to stakeholders, supporting strategic and operational decision-making. This project significantly advanced Tereos’ capabilities in leveraging data for agricultural optimization, delivering actionable insights with tangible economic impact.
Senior Data Scientist
Jan 2020 - Mar 2023
3 years 3 months
Région de Paris, France
My team is tasked to bring AI and ML into manufacturing 4.0 to generate more revenue streams and reduce costs. My tasks involve to take in charge the whole lifecycle of the projects, which includes : - As the tech lead, I made the cloud architecture of the projects and data pipelines to mutualize use cases and ease deployment on Azure Cloud. - Studied, analyzed,developed and deployed on the cloud recommendation systems and ML algorithms to improve the production of the plants and reduce energy costs. - Validate the solutions with the business. - Industrialize the solution and deploy the solution in the cloud (Azure) Technologies : Python / SciKit Learn / Numpy / Docker/ Kubernetes / SQL and NOSQL Databases / Queues (Celery) / Azure Blob Storage / Message brokers Azure Event Hub/ PowerBI / Azure Devops (CI/CD). Methodology : Agile Scrum / Devops
Data Scientist
Total
Mar 2018 - Dec 2019
1 year 10 months
Paris Area, France
Developing analytical methods for managing unconventional and conventionnal assets: Forecasting on time series using machine learning/ Deep Learning to help management make quicker and better decisions. My main tasks were: - Exploring different machine learning models and deep learning architectures to deal with each use case (geological data/ time series data/ mix of dynamic and static data ...) - Defining new metrics to assess the performance of our models against a physical one. Then providing a statistical study on each case to justify the approach and validate results. - Exposing the models with REST API. - Developing web interface to interact with the models and visualise the results. - Studying different deployment solutions. - Deploy the model into production for worldwide access by managers / field engineers. Tools: - Python : Pandas , Numpy, Sci kit, Keras, Tensorboard, Flask, Apache Airflow. - Web: Vanilla JS / Html/ CSS/ Bootsrap - PowerBI. - Project management: Jira (agil Kanban) / BitBucket (git) - Deployment: Jenkins
Research Assistant
École Polytechnique
Oct 2018 - Mar 2019
6 months
Paris Area, France
I am interested in numerical black-box optimization where derivatives are not available. I am focusing especially on adaptive stochastic black-box optimization algorithms or randomized search methods like Evolution Strategies whose probably most prominent algorithm is the Covariance Matrix Adaptation Evolution Strategy (CMA-ES).
Artificial Intelligence Engineer
Epilert
Oct 2017 - Mar 2018
6 months
Paris Area, France
Epilert is a wearable device that detects and monitor epilepsy seizures, it is connected to a mobile app and doctor platform in real-time run by an artificial intelligence which supervised by the doctors. It sends notifications to parent and caregiver during the crisis.
Engineer Internship
Télécom ParisTech
Apr 2017 - Oct 2017
7 months
Paris Area, France
The internship was about "Implementation of accelerated coordinate descent method for the Lasso problem " As a part of a team I helped in: - Study of convex optimization tools such as gradient calculations, proximal operators, the notion of dual problem. - Implementation of the N-dimensional gradient checking - Implementation of an accelerated coordinate descent method for the Lasso problem in Machine Learning. - Using the restart technique - Writing an effective code in terms of calculation resources - Deployment of the algorithm within the Scikit-Learn library. Tools: - Python / Cython
Education
Université Paris Cité
Master's degree
2016 - 2017
1 year 2 months
Ecole Nationale d'Ingénieurs de Tunis
Engineer’s Degree
2014 - 2017
3 years 2 months
Preparatory Institute for Engineering Studies of Monastir (IPEIM)
Engineer’s Degree
2012 - 2014
2 years 2 months
Licenses & certifications
Machine Learning Engineering for Production (MLOps)
Issued: Aug 2022
Sequence Models
Issued: May 2019
Deep Learning Specialization
Issued: May 2019
Structuring Machine Learning Projects
Issued: Apr 2019
Neural Networks and Deep Learning
Issued: Apr 2019
Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization
Issued: Apr 2019
Convolutional Neural Networks
Issued: Apr 2019
Participation in ACM TCPC Tunisian Collegiate Programming Contest 2016
Issued: Mar 2016