Nicolas Peltre

• Engineered and industrialized end-to-end machine learning pipelines for sales forecast generation, implementing monitoring, CI/CD, and integration testing. • Designed and implemented database structures to support pipeline outputs and business needs. • Enhanced Data Science and Analytics team efficiency by optimizing the codebase.

• Optimized, deployed and monitored real time Computer Vision models on Kubernetes, serving more than 1B requests/month, up to 300 RPS - 99.9% uptime. • Project owner and main contributor of several Kubeflow pipelines to download and generate millions of image embeddings with any models, serving them on a vector database. • Improved MLops practices, to automatically optimize model serving by measuring tradeoff between quantization, serving parameters, and deviation. • Developed a MVP RAG system, leveraging internal documentation and integrating it with Slack. • Enhanced knowledge sharing and collaborative decision making processes within the team.

• Engineered and deployed a marketing multi-touch attribution model (Markov chains) – enabled a 15% digital marketing budget shift. • Developed and owned an optimization tool for SEA campaigns through -demand/offer analysis – improved marketing ROI by 10%.

• Designed and deployed an NLP classification solution processing 15k emails/day – improving response time by 40% through automated routing and responses. • Developed and deployed a time serie drift detection model for web marketing data quality – reducing detection time by 80%.