Amine Boujida

Took a career break due to urgent personal reasons abroad. During this time, I'm working independently as a contractor on a few short-term remote projects while also focusing on personal initiatives and learning. Registered as auto-entrepreneur in France. Responsibilities include :

• Provide guidance on modeling strategy

• Advise early-stage teams on ML system architecture, Spark - Python - SQL - Pytorch

As a Data Scientist at Amazon Ads, I develop and implement advanced data-driven strategies and ML algorithms to enhance ad relevance for both in-store and offsite placements, contributing to a platform serving millions of advertisers and reaching hundreds of millions of customers globally. My role includes leading team initiatives and influencing broader decisions on targeting strategies. Project Highlights

• Contextual Targeting on Twitch: Developed privacy-preserving contextual targeting solutions for Twitch, introducing audio-based content analysis methods and new statistical frameworks for con tent relevance using Monte Carlo permutation tests. Built embedding-based tools to track content stability across streaming sessions, showing strong correlations with ad performance metrics and en abling effective targeting without user identification. Led the entire process from research through to production, achieving a 14% revenue increase in controlled experiments.

• Audience Targeting Optimization: Identified and addressed performance bottlenecks in audience tar geting channels , coming mainly from imbalance in the demand distribution and attendance bias. Implementing a recommender system that boosted audience engagement and contributing to a 12% increase in CTR.

• Content Analysis and Crawling Optimization: Led initiatives to analyze and optimize web crawling strategies, identifying significant opportunities for improvement. Successfully implemented a con trastive learning approach to exploit these opportunities

• Mentorship and Collaboration: Mentoredjunior team members and interns Co-Authored internal papers AWS - Spark - Java - Pytorch - Recommender Systems - Contrastive learning - 2 Tower Models

• Audio signal processing - LLMs

Engie, a market leader in residential electricity, serves over 7 million households. I leveraged smart meter and customer datasets to enhance customer insights and optimize energy consumption through web and mobile services. As a Data Scientist on a multidisciplinary team of 11, my responsibilities included data analytics, designing modeling solutions, and supporting production deployments.

• Conducted several studies : Assessing Covid-19 impact on household energy consumption , identifying a positive correlation between churn rates and end of the year regularization bill

• Developed a multi-granularity electricity consumption forecaster (Yearly , monthly , daily , hourly ). +21% coverage and -54% in the regularization bill. Self supervised approach to enhance the model's resilience against data anomalies .

• Developed a simple yet effective client segmentation approach

• Hypothesis testing during new products launches C3AI - Spark - TensorFlow - Time Series modelling - Causal inference - Signal processing - Hypothesis testing - AWS