Pierre-Emmanuel Diot

Built and scaled a "Visual Discovery" mobile app for secondhand fashion, growing from concept to 100,000 users. Defined the technical strategy and deployed the Data & ML infrastructure connecting Pinterest inspirations to Vinted's inventory. ＞ Multimodal Search Engine: Designed a Retrieval & Ranking system using Vision-Language models (FashionCLIP) and vector search (Pinecone) to match raw images against millions of listings. ＞ Data Engineering: Architected ETL pipelines on GCP (BigQuery) for the ingestion, cleaning, and indexing of over 5 million unique items in near real-time. ＞ Backend & API: Designed a REST API orchestrating vector search on Vinted's catalog, Pinterest social integration, and real-time personalized recommendation algorithms. ＞ Web & Mobile Development: Launched the initial web platform (React) followed by a migration to a native mobile experience (React Native) to maximize user retention and engagement.

Developed AI and Data solutions to automate financial product analysis and optimize conversion funnels. ＞ GenAI & RAG: Built LLM-based pipelines (RAG) to extract structured data (JSON) from unstructured financial documents (PDFs) and web pages. ＞ Modern Data Stack: Engineered a central Data Warehouse unifying marketing spend (Google Ads), leads (Typeform), and revenue data for precise ROI attribution. ＞ Lead Scoring: Developed a predictive model based on user navigation patterns to identify and prioritize high-value users.

R&D initiative at Epsilon (Publicis Groupe) to industrialize marketing performance measurement workflows. ＞ MMM Standardization: Development of an R-based framework to automate and standardize econometric modeling processes. ＞ Feature Engineering: Implementation of algorithms for media variable transformations (Adstock, Diminishing Returns) and hyperparameter optimization. ＞ Decision Support: Creation of simulation modules (response curves) to estimate incremental channel contribution (TV, Digital) and maximize marketing ROI.

Applied statistical and econometric methods to model biometric and sports-related risks. ＞ Data Visualization (COVID-19): Development of an automated dashboard for real-time pandemic monitoring and portfolio impact assessment. ＞ Econometric Modeling: Conducted a statistical study quantifying the correlation between physical activity and mortality rates (Life Risk). ＞ Data Engineering (Sports): Implementation of an end-to-end Web Scraping and data cleaning pipeline to build a proprietary dataset on professional football player injuries.