Tristan Marechaux

Tristan Maréchaux is a senior machine learning engineer with extensive experience developing and optimizing machine learning models. He currently holds the role of Senior Machine Learning Engineer at Doctrine, a Paris-based company, where he implements AI solutions to improve the company's products and services. Previously, Tristan worked at Leboncoin.fr as Staff Machine Learning Engineer, where he played a key role in improving collaborative filtering and semantic search algorithms, while ensuring the implementation of good data engineering practices.

During his time at Leboncoin, Tristan demonstrated his skills by replacing the approximate nearest-neighbor search algorithm with state-of-the-art techniques (ScaNN), which led to a significant reduction in operating costs and improved latency. He also designed and maintained a semantic search re-ranking algorithm using BERT and triplet loss, enabling the deployed model to serve 1,000 requests per second with 20ms latency.

Before focusing on AI, Tristan worked as a software engineer at Walnut Algorithms, where he contributed to developing and maintaining real-time trading software using machine learning techniques. His role included scaling the software using microservices, improving deployment and availability with Docker and Kubernetes, and integrating machine learning components into the automated trading software.