Stamp

Compensation

Full Description

About Stamp We believe that Human Intelligence is the most valuable currency - one that should only be spent on solving the most challenging and important problems. That’s why we’re building Stamp, the first AI Secretary. Stamp handles your emails and calendar for you, even when you’re not using it, so you can focus on the problems that matter most. Stamp is the first product of its kind, capable of handling complex user workflows (drafting emails, searching the web, cleaning up mailboxes) via its Agent when a user is online, while also automatically summarizing, categorizing, and drafting replies to emails while a user is offline. We’re a rapidly growing, fast moving company, with single minded focus on elevating our users’ productivity through cutting edge technology. About the Role As a founding machine learning engineer, you’ll be working to help improve the performance of Stamp’s models and agents, working on the cutting edge of the field to deliver innovative user experiences. Moreover, you’ll have the opportunity to: Own the model architecting, training, and evaluation process for custom built transformer models Work on our evaluation metrics and datasets for custom built models and agentic workflows Construct new agentic workflows powered by a custom built agentic framework We’re looking for engineers who: Have direct experience training ML models from scratch and/or building full stack AI applications Have worked with ML training frameworks like PyTorch, especially in multi-GPU and/or distributed training environments Love agonizing over model performance and highly optimizing them for both speed and performance Job Requirements Significant experience working with PyTorch and training Transformer-based models Experience building accurate agentic systems US Citizen, able to work in person in San Francisco