- Helping modeling client-side problems into datascience problems - Working with the team to choose and evaluate Machine Learning / Deep Learning Algorithms - Improve Data Workflow - Leading tech team for Modeling and training Deep Neural Networks
Teaching to Master students the art of data science, AI and can those leverage business.
Context: Students are from management and marketing field,
Goals: - My role was to give them the theorical and practical knowledge of data science and AI to give them the ability to understand, setup and even make a data science project in a professional ecosystem.
- Give the student the knowledge to interact with the data science team and to understand all the process of a data science / AI project.
Level: Students are in their last year of Master degree
- Out Of Stock & Abandonned Card Recommendation Algorithm: Recommendation Algorithm using image style content of an item to extract the "style" and match it to other items. Fashion Industry specific using the style of an item as the dna of this item
- Media for Branding Recommendation Algorithm: Using web log visit to extract taste for media ads branding. Recommendation of Category using web log visit
As a Lead Datascientist in freelance I succefully helped the team to define and build two main projects in machine learning apply to marketing.
Project 1: Score the likelihood of a customer to buy a product
---- Context: - 6 year of user activity logged in a DataLake (hdfs) - 1 year of user activity in the website logged in google analytics - Cross Domain products ( Food, Home, Jewelry .... ) - About 300k active users per year
---- Goals: - Increase the cross selling by targeting the wrigth user for the wrigth product when a user didn't buy the kind of product (e.g. Selling chocolate when the user only buy jewelry) - Be able to define the value of the coupon for each customer ( higher is the attraction lower is the value of reduction ) - Understand the behavior of customers
Solutions: - A model that score the probability of a user to buy a type of product next month knowing his past (types of products are defined by the marketing team) - A model that map users and products into a semantic space that models the behavior and interactions between user / user and user / item
Work Done : - Lead the team to success - Formation of the data science team - Extraction of data in the DataLake (Spark 2) - Extraction of data in google analytics (Python + Google Analytics) - Feature Cleaning and Feature Extraction, Feature Engineering - hyperparameter and parameter optimization + Model selection - 1 classification model with probability outputs (H2O + Sklearn) - 1 recommendation model with score output + mapping (TensorFlow) Deep recommender systems
Keywords : collaborative filtering, Google Analytics Api, Hive feature extraction, Deep Feature Extraction
Project 2: Customer Segmentation on buying behaviors for Food Products
- Deep Learning: Research and implementation of deep learning algorithms for Natural Language Processing (LSTM, CNN, W2V, C2V and more) (Python, Theano, TensorFlow)
- Chatbots: Implement and research new chatbots structures. Clustaar was a pioneer of chatbots technology and my role was to help them find new paths to build their system.
Implementation of clustering algorithms for keyword clustering Big Data (Scala, Spark), using this to extract intents and entity in natural language sentences, automation of intention extraction.
Consulting missions in helping compagnies to exploit their user's data. Data extraction, Data mining, Data agregation, user segmentation / clustering, behavior analysis and prediction (Python, Sklearn, Numpy, MongoDB, Sql).
Internship student Mentor : Recruit, Teach and Manage an 6 month Intership Student in the field of Object Recognition in Images by using Deep Learning for client Mappy (french version of google maps). (Python, Tensorflow, C++)
As a Lead Data Strategist in this emailing marketing compagny I built from scratch the infrastructure.
Work Done: - Coordinate the technical team and the interactions with the commercial team. - Define the scope of project to enhance business with data - Define the digital strategy roadmap - Build the infrastructure of servers that send mails ( more than 300k emails send by day ), the infrastructure was built in a way that optimize speed, security and reputation of ips. - Develop a software to manage emails segments, work on commercial email, Bulk emailing, statistics reports. - Addons for IP rotation, IP reputation monitor, Speed per IP based on reputation and past actions. - Develop Business Intelligence tools for decision making. - Build from scratch Machine Learning tools to segment customers by their behavior.
Talent sourcing specialist de la plateforme Pentalog, SkillValue s’appuie sur un pool de +400 000 Experts Tech & Marketing, dont +15 000 freelancers. Évaluations de compétences, missions, offres d’emploi : les consultants SkillValue vous aident à accélérer votre carrière.
Restez informé de l'actualité tech avec la newsletter Pentalog
À propos de Pentalog
Pentalog est une plateforme de services IT qui accompagne les entreprises dans leur production digitale et leur transformation en fournissant des ingénieurs et développeurs de classe mondiale. Ses 16 Delivery Centers et Agences dans le monde comptent plus de 1600 ingénieurs, développeurs, product designers et spécialistes marketing.