Researcher

Trace is building data infrastructure that enables robots to learn from real-world data and supports the development of physical AI systems. The Researcher will conduct frontier research on robot learning and data quality, run experiments, develop evaluation methodologies, and collaborate with engineering and data teams to translate findings into pipeline and product improvements.

Responsibilities:

• Design and run experiments on training robotics, vision, and embodied AI models using Trace's real-world data
• Investigate how data quality, structure, and diversity affect downstream model performance, and turn those findings into concrete recommendations for what we capture
• Build and iterate on evaluation methodology so we know whether a model, or a dataset, is actually getting better
• Stay close to the literature on robot learning and translate what's relevant into work we can actually ship
• Work directly with the engineering and data teams so research findings turn into pipeline and product changes