Coveo provides search and recommendations services to more than one thousand clients around the world. The machine learning platform and algorithms must support multiple use cases in 29 languages with clients having datasets of various sizes and specific vocabulary. This context brings multiple challenges when developing NLP algorithms that should work in 29 languages or more, adapt automatically to thousands of datasets and provide advanced capabilities in multiple fields such as question answering, query understanding, support case descriptions to document documents similarity, and knowledge extraction and retrieval. This talk will present Coveo, it’s main use cases and look forward and present challenges that could be tackled in future research projects.