Erich Elsen

As a Principal Research Scientist, I focus on advancing the state-of-the-art in AI and machine learning, contributing to Databricks' core data and AI platform capabilities. My work involves high-impact research projects aimed at improving model efficiency, scalability, and performance for enterprise applications. I collaborate with engineering teams to transition cutting-edge research into production features.

Co-founded Adept, an enterprise AI startup focused on building general intelligence capable of performing a wide range of tasks for knowledge workers. I was instrumental in developing the foundational models and infrastructure that powered Adept's early agent tools. The company was acquired by Amazon in June 2024, marking a significant exit.

Contributed to DeepMind's mission of solving intelligence, focusing on large-scale machine learning systems and optimization techniques. My research involved developing efficient training methods for complex neural networks, leading to several high-impact publications in top-tier AI conferences. I specialized in performance optimization for deep learning models.

Led the High Performance Computing (HPC) group at Baidu's Silicon Valley AI Lab, focusing on scaling deep learning training and inference. I developed novel techniques for training Recurrent Neural Networks (RNNs) at scale, significantly reducing training time and improving model accuracy. Presented findings at major industry conferences like MLconf.