Florent Lefort

I designed and implemented a centralized Job Offers Hub and an interactive dashboard to provide a unified, reliable, and actionable view of job posting activity across multiple fragmented Data Lake sources. The goal was to eliminate manual data reconciliation and complex SQL queries previously required by Product and Employer Branding teams, and to enable self-service access to key performance indicators.

I built the Hub by aggregating and reconciling data from multiple tables (job postings, alerts, applications, and hiring declarations) into a single structured data model. On top of this foundation, I defined and calculated key KPIs such as the number of active postings, newly published offers, and application volumes. I then developed an interactive dashboard with dynamic filters to enable real-time exploration of these metrics and support daily operational monitoring.

I also supported functional users in adopting the solution and participated in cross-team coordination to ensure consistency across interconnected data hubs. The final solution became a shared reference for product managers, significantly reducing time spent on data extraction and enabling real-time detection of anomalies in job distribution activity.

I designed and deployed a machine learning-based decision support system to automate and explain the validation or rejection of banking operations (e.g., large transfers and cheque deposits) handled daily by BPCE advisors. The objective was to improve decision speed, consistency, and reliability while maintaining full interpretability for end users.

I built the training dataset by extracting and structuring data from the SGBD based on business-defined criteria, and engineered relevant features to improve predictive performance. I then developed a classification model to predict whether each transaction should be validated or rejected, and defined variable-specific thresholds to support decision logic. To ensure transparency, I implemented explainability mechanisms by computing and exposing the top contributing factors for each prediction.

Finally, I designed a monitoring dashboard to track model outputs and detect performance drift over time. The solution was deployed in production and is now used daily by banking advisors, providing instant recommendations with clear explanations, significantly improving trust in the system and supporting faster, more consistent decision-making.

I designed and implemented a marketing mix modeling solution for Carrefour to measure the impact of advertising investments on sales performance and optimize media budget allocation. The goal was to quantify the incremental revenue generated by advertising campaigns and provide marketing teams with a data-driven decision-support tool for future budget planning.

I built a centralized data mart by aggregating sales data across product categories and store departments from relational databases. I engineered promotional and media investment indicators to isolate the effect of advertising campaigns on sales and developed explainable linear models to quantify the contribution of TV, radio, and promotional activities to revenue generation.

I also created an interactive dashboard that enables stakeholders to visualize model outputs and simulate budget-allocation scenarios across media channels. The solution showed that advertising campaigns generated approximately 5% of total sales, representing several million euros in attributable revenue, and helped marketing teams optimize future campaign ROI.