Marie-Noël Chanthavong

Data Scientist at Riverty

Fraud Modelling: Collaborate with Risk team to define problem statement and model metrics. Build and deploy ML serving endpoints (FastAPI) near real-time models (tree based) to detect fraud cases. Simulate business metrics given different scenarios. Achievements: Prevent 10-40% losses by respecting acceptance rate of merchants. Product returns prediction: PoC of classification models applicable across multiple merchants to forecast the likelihood of returned orders within 3 months. Feature Engineering repository: Key maintainer of a mono repository for feature engineering, reducing code redundancy and deployment time for ML projects.

Data Scientist at LMK group: Godtlevert

Time-Series Forecasting: Predict total orders and number of meal-kits for short-term/long-term (FB Prophet, ensembling model-PyCaret) in 11 weeks ahead. Improved model: 2x faster training time with same accuracy; monitoring models; analyzing total orders by geographical zones; Automating Stored Procedures via Retool UI for forecasting model; NLP: data modeling for customer communications channels, sentiment analysis, text classification of menu rating comments; Insight analysis using Bisnode data; customer segmentation; Picking optimization; Tableau dashboards for forecast performance and onboarding reports.

Research Assistant at National Veterinary School of Toulouse - INREA, EPIDEC

Genomics analysis of the respiratory microbiota: characterization of the air microbiota between two different environments; flat-files processing; using PLS-DA to reduce variables; ANOVA tests. Milk and Dairy in India: regression models to explain determinants of dairy dynamics (1995-2015) from structural context variables and geospatial analysis.