B

Build AI

Public-benefit data hyperscaler for physical AI

Member of Technical Staff

AI ResearchConfirmed live in the last 24 hours
Full-timeNo salary listed

About the Job

Requirements

  • The candidate must have trained real models and be able to defend a research bet.
  • The candidate must be proficient in Python and a deep learning framework such as PyTorch or an equivalent.
  • The candidate must want to work on in-the-wild physical data rather than only academic splits.
  • The candidate must be able to move between research questions and engineering to complete an experiment.
  • The candidate must be able to operate independently on an ambiguous, high-impact direction.

Responsibilities

  • Run research bets on Build data by training, evaluating, and ablating models, then shipping what works.
  • Design and train models on in-the-wild video and physical-labor data, determine what is learned as the data scales, and identify what is still missing.
  • Build needed training and data paths, including PyTorch, distributed training, and dataset curation.
  • Work with the Head of Dataset & Quality and Evals Lead so results influence data collection and evaluations, not only papers.
  • Cover research and engineering as needed in a team of fewer than 30.
  • Design evaluation frameworks for research bets that assess generalization, not only training curves.

Desired Qualifications

  • Experience with video, world models, robotics, or multimodal training.
  • Experience with large-scale pretraining, dataset curation, or distributed GPU training.
  • Familiarity with egocentric, first-person, or in-the-wild video.
  • Published work or shipped models that changed what a lab did next.

About the Company

Build AI

Public-benefit data hyperscaler for physical AI

Build AI runs as a public benefit corporation that acts as a data hyperscaler for physical artificial intelligence, handling hardware, data collection, and model training in one end-to-end pipeline. It collects an egocentric dataset of physical labor by equipping thousands of workers with head-mounted cameras in real-world settings, then provides this data to researchers and corporate robotics teams to train physical AI models. The company differentiates itself by being vertically integrated—owning the hardware, data collection, and training process—and by focusing on large-scale, real-world data rather than relying on third-party sources. Its goal is to accelerate the development of robots that can perform manual tasks in everyday human environments, unlocking human potential.

Company size
11-50
Company stage
Not listed
Total funding
Not listed
Headquarters
San Francisco, CA
Founded
2025

Recently Posted Jobs

Explore more opportunities at Build AI

See more jobs