Data Scientist

• Conceptualized, developed, and iterated upon a highly effective Machine Learning-based recommendation algorithm, remarkably enhancing the relevance and volume of suggested partners by a notable 76% by incorporating over 40 features based on CRM data and Graph metrics (NetworkX)
• Created a LinkedIn industry mapping algorithm combining embedding technologies (Bert, Sentence-Transformers, OpenAI) and an efficient supervised neural network model, to significantly propel industry normalization accuracy from an initial 72% to an impressive 98%
• Engineered a streamlined automated Python script to seamlessly generate over 50+ meticulously refined aggregated lists of new prospects to
premium users, catalyzing a substantial average ROI surge of up to 300%
• Increased the data quality by 80% on the Analytics page by conducting rigorous statistical tests (1M+ data points) and integrating a robust BoxCox algorithm to eliminate outliers with a 95% confidence level
• Implemented in production 20+ Python unit tests (Pytest framework) of the previous solution, ensuring 100% stability in data computation and comprehensive coverage of use cases to prevent future bugs
• Investigated multiple LLM models, including GPT-3.5, to enrich missing CRM data by 33%

• Led the “Care Advisor Project” for Van Cleef & Arpels, orchestrating a successful 6-month implementation of a multilingual Natural Language Understanding (NLU)-based virtual assistant with consistent client interactions
• Achieved a notable 80% decrease in service request triggering time by deploying the Care Advisor on VCA’s website, alongside a remarkable 62% upswing in customer satisfaction verified through VCA’s survey
• Developed and deployed a robust Python-based correction system for the NLU-based assistant, incorporating 5K+ use cases. Expertly integrated it into the product’s back-office, resulting in an 8-hour/week reduction in operational costs
• Engineered weekly KPI reports with Google Looker Studio for Dior Visual Merchandising and drive a 40% improvement in global boutique visual merchandising effectiveness as reported by boutique managers
• Authored comprehensive specs for “Kellogg’s Wellbeing Project” and created over 100 JIRA tickets, effectively guiding the development team in a 25% reduction in delivery time and a 65% decrease in bug-fixing cycles as measured by JIRA Control Chart