Data Scientist

• Jan 2022 - Present

Data Scientist Freelance

• Nov 2020 - Jan 2022

Working on projects ranging from data analysis and reporting to the development of POC involving Deep Learning (NLP and CV). Working on projects ranging from data analysis and reporting to the development of POC involving Deep Learning (NLP and CV).

Science to data science

• Aug 2020 - Sep 2020

S2DS is a 5-week data science bootcamp helping analytical PhDs transition from academia to industry. Fellows accepted into the program learn via hands-on projects, and work in groups to complete real-life big data problems. The curriculum also includes 30 hours of lectures on topics such as professional development and business skills, designed to help students achieve their goals after the program. During this program, I worked on a NLP project for the partner company Siisti.ai, along with 4 other team members. Our work involved remote team work using the Agile methodology, daily scrum meetings and weekly reports to our business partner. The project was concluded by handing over a working prototype of our product to our partner and by giving a presentation to the entire S2DS cohort. Our presentation was shortlisted as one of the three best presentations of the entire program. Show less

Data Science Competition

• Mar 2020 - Apr 2020

Bike sharing demand (Kaggle competition) : Top 5% best Machine Learning Algorithm The goal of this competition is to predict the demand for a bike sharing program (regression task). The solution I developed involved: - Exploratory Data Analysis (EDA) - Data cleaning and Data selection / Data engineering (creating new features) - Data preparation for ML models and model selection - Hyperparameter tuning of selected models (Gradient Tree Boosting, Random Forest, Neural Network) Hyperparameter tuning was realized through a K-fold cross-validation method and the best model was selected using a test set held out for this purpose. Show less

Data Scientist

• Sep 2019 - Mar 2020

Working on various case studies and self-teaching data science and machine learning. Case studies : - Black Friday sales: Perform exploratory data analysis to improve customer behavior understanding. Import, clean and prepare the data for machine learning algorithms. Develop algorithms Random Forest, Gradient Boosting, Neural Network) to predict future sales given a pair (customer profile / product type). Perform hyperparameter tuning to achieve better performances. Stack all three algorithms to obtain better robustness and better accuracy. Learning material: - Machine Learning course (Coursera, by Andrew Ng) - Data Science From Scratch (O'Reilly) - Hands-on Machine Learning with Scikit-learn, Keras and Tensorflow (O'Reilly) - The element of statistical learning (Springer) Show less

Chercheur postdoctoral

• May 2017 - Jul 2019

Electric dipole moments of elementary particles are of great interest in modern Physics to search for T-violating processes that could help us understand the matter/anti-matter asymmetry in our Universe. The neutron EDM has been studied for over 50 years with a sensitivity improvement of 7 orders of magnitudes and is still consistent with the Standard Model predictions. As part of the nEDM collaboration, I have worked on the analysis of the most sensitive neutron EDM data ever taken. I am also in charge of starting from scratch a new precision measurement of the mercury magnetic moment which will be a central piece of the next generation EDM experiment, n2EDM. - Statistical analysis / Systematic studies (C++/Python/ROOT/GitLab) - Multivariate regression - Time series analysis - Vacuum technologies - Laser spectroscopy - Optically pumped magnetometers Show less

Chercheur postdoctoral

• Jun 2013 - Aug 2016

Correlation coefficients in nuclear beta decay are interesting probes for new physics such as tensor-/scalar-like interactions and T-violating interactions. The beta-neutrino angular correlation in the Helium-6 beta decay is sensitive to new tensor-like contributions to the weak interaction. As a postdoctoral researcher for Argonne National Laboratory, I worked on the He-6 experiment at the University of Washington on developing the experimental setup for this precision experiment. I took part in the development of the Helium-6 production source from the nuclear reaction of deuterium and lithium, the laser cooling and trapping apparatus and parts of the detection system. As the first data became available, I also got involved in the data analysis and the study of systematic effects impacting our measurement. - Helium-6 production source design - Laser cooling and trapping - Data analysis (C++/ROOT) - Vacuum technologies Show less

Chercheur doctorant

• Oct 2009 - Nov 2012

Ion-atom interactions play an important role in areas ranging from tumor treatment to plasma physics and astrophysics. Using ultra-cold atomic targets, provided by a magneto-optical trap, and low energy ions allows to precisely study these interactions and provide stringent tests of theoretical calculations. During my PhD, I worked on the simulation of the electrostatic recoil-ion spectrometer, the development of the experiment, I took and analyzed the data and also got involved in the development of the theoretical calculations. - C++ (ROOT) - Data analysis - Electrostatic field, particle trajectories simulations (SIMION) - Laser cooling and trapping (Magneto-optical trap) - Vacuum technologies - Ion detection - Signal processing - Theoretical physics Show less

Lecturer

• Sep 2011 - Jun 2012

Teaching applied physics to undergrad students at the University of Caen (Mechanics, waves, electronics, optics). Teaching applied nuclear physics to engineer students from the ESIX school (Alpha, beta, gamma and X-ray spectroscopy). Teaching applied physics to undergrad students at the University of Caen (Mechanics, waves, electronics, optics). Teaching applied nuclear physics to engineer students from the ESIX school (Alpha, beta, gamma and X-ray spectroscopy).