Dave Farris

Dave Farris is a principal research engineer at NVIDIA on the Nemotron data, scaling and evaluation team. He develops evaluation methods that predict a model’s capabilities from early training signals and uses those predictions to guide data and training decisions. He is one of the creators of SWE-Serve, a benchmark that tests AI coding agents on the engineering behind LLM inference serving. Before NVIDIA, he worked on speech recognition at AWS, from model development through production deployment.