Baptiste - Consultant PYTHON

Data scientist (contract).

Supply Chain project. Development of a hierarchical sales forecast model at different levels
of geographics, time and products granularities. States-of-the-arts models (recurrent neural
networks). Intensive use of Python, Tensorflow, Dataiku.

Director : E. Ducros, ********

Responsible for the Machine Learning courses (Master 2 level). Mathematical theory. Linear and nonlinear regression, classification and clustering algorithms. State of the art deep learning models (cnn, lstm, siamese, attention). Python, Keras.

Study, implementation and evaluation of state-of-the-art deep learning models (cnn,
siamese, lstm, attentive networks, pointers networks) for text data (NLP solutions for
conversational agents system problematics such as question-answering, intention
disambiguation and duplicated questions detection) and images data (segmentation and
classification problematics). Intensive use of TensorFlow, Cuda, Python and R.

Deep learning scientist at Earthcube , Paris
September 2017 - May 2018 (9 months)
Contact: R. Allioux, ********
Study, implementation and evaluation of state-of-the-art deep learning (siamese, cnn)
models for satellites georeferenced images data (segmentation and classification for a
monitoring system) using Keras, Python, Qgis and R.

Visiting PhD at National Institute of Informatics

Machine learning and predictive models in heterogeneous data. I have proposed to integrate
a priori qualitative knowledge in structural-based models for ranking experts (individual with
tacit knowledges in a particular field of expertise) in digital libraries. Contributions have been
published in two leading computer science international conferences and associated results
have been presented in Suzhou (China) and Porto (Portugal).

Contacts: P. Gallinari, ******** , and L. Denoyer, ********

Optimisation of predictive models in Information Retrieval. I have proposed to combine
sentiment analysis and information retrieval models for ranking web reviews using both
relevance and polarity criterions.

Data Scientist

Contact: C. Balagué, ********
In a five members team working on the development of a CRM software for a French leading
actor in data marketing, I was responsible for the data analysis part. I have proposed a new
customers engagement metric by crossing social data (from individuals behavior on
Facebook pages) and transactional data (from customers' shopping cart). Proposed work
was implemented in the platform and published in the 36th Marketing Science Conference.

Data Analyst intern at IdeasVoice

Customers segmentation and classification from advertising campaigns data using R.

Data Analyst intern at IdeasVoice , Paris

Contact: Patricia Egard, ********
Customers segmentation and classification from advertising campaigns data using R.