Senior Applied Scientist
CurrentDeep learning model development using state-of-the-art NLP architectures. This includes data collection, curation, model training, evaluation, inference optimization, flighting & deployment of large scale models. More details about the model are as follows1) Content understanding model based on encoder-only architectures (BERT like) for tasks like article interestingness, evergreen & location-independence. These are multi-task models trained simultaneously on all these tasks and then distilled to a compressed student model meeting inference requirements.2) Personalized Neural Recommendation model ( both monolingual & universal ) built using semantic representation of documents, user profile based on historical user interaction & other contextual features. They are trained on objectives like predicting implicit user feedbacks ( click and read-time ) on news article. 3) Article summarization model based on decoder-only (GPT like) architectures for tasks like eye-catchy headline & search query generation.