Paul HERVET

• Scoped and deployed to production an AI assistant that helps travel advisors reorient prospects to better-fitting destinations, addressing ~12k yearly unconverted quotes (€5k avg margin/trip) • Architected a multi-stage pipeline: offline modular LLM extraction of proprietary data into structured features, then online conversational intent qualification → deterministic candidate filtering and ranking → LLM-generated personalized sales arguments • Built end-to-end evaluation framework: curated offline evaluation sets, iterations via expert and user testing • Optimized cost, latency, and reliability: right-sized model selection, parallel calls, typed outputs • Embedded travel experts' know-how in the product through weekly working sessions • Drove adoption: trained 80 advisors in batches, enabled power users

• Developed end-to-end data science applications, from design to production deployment, such as a pricing recommendation engine combining demand forecasting and cancellation prediction in the hospitality industry • Coached both technical and business users across pre- and post-sales to accelerate platform adoption and use case delivery, on key topics such as production deployment, machine learning, and tools integration • Delivered product deep dives to technical audiences and value-driven discovery demos to decision-makers • Scoped and led delivery of implementation engagements across 40+ accounts in Southern Europe, partnering with sales and customer success from pre-sales to rollout • Managed and grew a team of up to six data scientists and an engagement manager, leading recruitment and onboarding, and authoring advanced technical internal training workshops