My profile and technic/scientific competencies
Who I am
My experience as a senior data scientist in an innovation and marketing consulting firm put me in resposibility of end to end project completion. From the understanding of our customers’ needs and the translation of their business problems into data driven actionable solutions, to the delivery and communication of the result to C-level management.
I developed a wide range of data driven marketing solution from sales or investments forecasting to social media mining and sentiment analysis thanks to NLP implementation.
I also put in place growth hacking strategy to enhance the consultants' productivity and generate tailor made databases out of web and social network public but unstructured information.
Academic and professional backgroud
After my graduation as Bachelor in Finance and risk management with a strong emphisis on quantitative analytics at the University of Montpellier and School of Economics, Management and Statistics of Bologna, I joined the UCL for a MSc in quantitative economics, econometrics and modelling.
I had the opportunity to put in action my academic background during three years as a data scientist and economist at the French Ministry of defence where I developed and test models to maximize assets disponibility under binding budget constraint.
I am therefore a nowadays 4 years experienced data scientist at a professional level with a strong academemic background (scientific publications and conferences) in statistical and economic modelling. My profile is thus oriented toward statistical sciences, but my professional experiences and passion pushed me to train myself to computer sciences and to be able to write production level code.
Professional background other than data science
During my early professional life, I was a duty personnel in the french armed forces. I served in the alpine troops but later I served in the military staff where I was responsible of the relationships between the armies and the parliament. Practically, I was answering the deputies and senators questions about the fleet availability and the actions put in place by the armies to improve it.
See my generic CV →Python main library for data manipulation and visualisation
Python main libraries for Machine learning and NLP
Main libraries for deep learning
Main libraries for web scraping
Other (Web dev microframework and other languages)
My competencies and added value are supported by many pillars
Scientific thinking
My training and first work experiences gave me the opportunity to publish in peer reviewed journals, aswell as participating to international conferences in econometrics. I also taught statistics and econometrics at univerty for undergraduate students. This gives me the advantage to never use models as a blackbox and a vast knowledge of model variety, allowing to find the most pertinent one to the problem under study.
Variety of real world problems tackled
I have worked in many fields ranging from industrial and military optimization to voting behaviour analysis, through real estate, new space economy or sport and healthcare. This pushed me to work with multiple type of data (georeferenced, time series, panel, truncated...) and to apply a wide variety of models. All supported by research projects or presitigious clients.
Project management
In charge of the full completion of projects I am implicated in helping the commercial forces to sell projects. I therefore judge of the feasibility of the project and design the technical solution associated. After the operational realization of the project, I am responsible of the result presentation to top level management. I am therefore experienced in defending the results and their implication to high level professionals, either in the public or private sector.
Entrepreneur spirit
I like things to move forward and to be a force of proposition to put them in motion. In my current company I was one of the main person in charge of developping our new data science offer toward our customers (marketing, innovation director in major aeronautics or agrofood companies).
Some projects examples !
Prediction of the investments in the US new space sector
- Objective: forecast of investments
- Data: collected via extensive scraping
- Methodology: Vector Error Correction Model
This project aims at predicting the strategy of investments in space economy from private actors in the US. The prediction takes advantage of the long run relationship between the investments and the quantity of launch made by US rockets. A battery of statistical tests (Granger causality, co-integration) oriented us toward a VECM modelling showing fairly good predictive accuracy at a 3 year distance.
Estimation of the Houses price jump after the implantation of new transport amenities
- Objective: predict the variation in house prices in Toulouse in response to new suburb construction
- Data: House selling national notary association database
- Methodology: Spatialized causal effect (Rubin) model - Data viz interactive mapping
This project estimates the effect of new transportation amenities (a new metro-line) on the price of houses located "nearby" a new metro entry. The estimation is based on a spatialized Rubin (causal effect) model. The test and control group vary with different radius size whose center is represented by a metro entry to test the effect of a new transportation amenity at different distances.
Predicting mortality probability 6 month after heart failure event
- Objective: predict mortality for patient after heart failure event based on biological and lifestyle data
- Data: Anonymized data of US set of clinics
- Methodology: Test of different classifiers - Logistics, ANN, ...
The model help to predict the mortality of patients six month after heart failure incident based on biological characteristics and lifestyle parameters. Multiple classifier models are tested. The aims resides more in writting ready for production scripts (that can be found in the github repo) than fine tuning the classifiers to optimize prediction, although the latter remains fairly good.
Creation of an estimator to uncover spatial data generating processes
- Objective: Econometric research
- No data as pure modelling
- Methodology: Use of the properties of LASSO estimators and High dimensional statistics to estimation of a spatial autoregressive process
This project is an advanced modelling statistical tools to help uncover hidden spatialized pattern into data generating processes. It uses the sparsity within the data generating processes to shrink small parameters to 0 and remove them from the model until the systems is identified and estimable.
Voting behavioural analysis
- Objective: analyze how votes for eliminated candidates transfers to remaining candidates in a second round
- Data: district level for 2012 french presidential elections
- Methodology: Non linear estimation of a transition matrix
Passionate about politics I was curious to understand how the votes from the extreme right voters would transfer once their candidate eliminated from first round. The project was aiming to give better understanding of the porosity of votes between classic right and extreme write in a context where extreme right leaders tend to make their opinion less extreme and more acceptables to median voters.
Productivity increase projects
- Objective: automating repetitive internet search to increase non-tech collaborators productivity
- Data: Open Web
- Methodology: Web scraping
The objective of this set of notebook is to automate repetitive task that consultant went through during their investigation phase for their respective projects and client. The notebook associated as example is just a sample of things that were made as each project is tailor made for each project.
