Seedify

OpenAI has shown us how it's done. The first milestone over the next 6 months, is to train and optimize our language models, with a target of a \u0026lt;1000ms latency and \u0026gt;20% suggestion acceptance rate for a curated list of autocompletion scenarios.\n\nWe'll start with 20B open-source, pretrained GPT neox, but will go back and forth between much smaller models for validation. We have enough data (millions of documents) and compute (16xA100) for training.\n\nYou'll be on the core team that builds the pipelines for training, inference and evaluation. \n\nThe first few weeks can be joined remotely. In the longterm, we will also have US offices and up to 8 weeks p.a. of remote work. \n\nApply if \n- you have some experience with evaluating models, optimizing inference pipelines \n- and a very strong research or coding background.