Bartosz Wróblewski

Bartosz Wróblewski holds the position of Member of Technical Staff and has made significant contributions to the field of machine learning. He co-authored the influential paper ‘Scaling Laws For Every Hyperparameter Via Cost-Aware HPO’, which explores the scalability of hyperparameters in relation to their cost-efficiency. Wróblewski also played a pivotal role in the development of CARBS, a cost-aware hyperparameter optimizer designed to enhance model performance through efficient tuning. His research includes demonstrating the substantial impact of hyperparameter tuning on model performance and reproducing the Chinchilla scaling laws for large language models (LLMs) using CARBS. Additionally, Wróblewski worked on a project that successfully addressed OpenAI’s ProcGen benchmark by tuning a simple baseline model, further showcasing his expertise in optimizing machine learning processes.