Anna Goldie

Anna Goldie is an AI researcher and entrepreneur whose work integrates machine learning, computer hardware, and natural language processing. Goldie earned a doctorate in computer science from the Stanford NLP Group and holds bachelor's degrees in computer science and linguistics, as well as a master's in electrical engineering and computer science from MIT. Her expertise spans systems, language, and learning, shaping a career focused on advancing artificial intelligence and chip design.

Goldie co-founded and led the ML for Systems team at Google Brain, where her research centered on deep reinforcement learning for computer systems, particularly chip design. Her work, in collaboration with Azalia Mirhoseini, resulted in a paper published in Nature that introduced a deep reinforcement learning system capable of designing chip layouts. This AI system approaches chip floorplanning as a sequential decision-making process, optimizing component placement and improving layout quality. The technology has been implemented in multiple generations of Google's Tensor Processing Units, datacenter CPUs, and chips used by Alphabet and external manufacturers.

Goldie has contributed to research on Constitutional AI and retrieval-augmented large language models at Anthropic and led large language model research on Gemini at DeepMind. Her publications include "Constitutional AI: Harmlessness from AI Feedback" and "RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval," covering topics such as AI safety, language modeling, and systems research. She has been recognized by MIT Technology Review as one of the 35 Innovators Under 35 and received a Best Paper Award at the ICML Workshop on Generating Natural Language.

Goldie and Mirhoseini co-founded Ricursive Intelligence, an AI chip development company focused on accelerating chip design through AI automation. The company has secured significant funding and aims to enable a self-improving cycle between AI and hardware, exemplified by AlphaChip, which was trained on TPUs and subsequently designed new generations of TPUs. Goldie has also delivered keynote presentations to international audiences, including Google Developer Day in China, where she addressed 10 million viewers in Mandarin.