AI Data Trainer
Outlier.AI is a subsidiary of Scale.AI whose aim is to provide labelled and functional ready to use data inorder to train AI models and chatbots belonging to consuming companies like META or Amazon. The tasks were centered generally on: - Data collection -Data labelling and ranking - Prompt generation - Response Review, Analysis and Comparism I worked on the Project Flamingo Bulba, Flamingo ranking and justification, Flamingo Eval and Flamingo Preference Multilingual. The target language used was french. The quality of the submitted work was based on the respect of the following criteria: - Accuracy - Completeness - Localisation - Fluency - Relevancy