Summary
RescueSight is hiring a Spatial Perception Intern to work on 3D vision, spatial AI, and robotics. The role involves developing systems that enable robots to reconstruct, understand, and remember environments, while implementing experiments and integrating successful ideas into core systems.
Responsibilities
- Streaming 3D reconstruction: Building detailed, persistent maps from video as it arrives, without access to future frames
- Gaussian splatting and neural representations: Getting sharper geometry, better novel views, and efficient updates from feed-forward models
- Geometry and long-horizon mapping: Improving depth and camera tracking, reducing drift, and handling revisited places and changing environments
- Spatial understanding and memory: Connecting what a robot sees to persistent objects, places, and relationships it can use later
- You’ll help decide what to try, implement it, run the experiments, figure out why something failed, and bring successful ideas into our core systems
- You’ll work directly with our founders and technical team, with backgrounds in Google DeepMind, Meta Reality Labs, Toyota/Honda Research, and computer vision + VLA/WAM research
Skills
- Strong Python and PyTorch skills
- Depth in a relevant area
- The ability to run careful experiments
- Open to undergraduates, master's and PhD students, new graduates, and early-career researchers and engineers
- A project you took seriously, a paper you reproduced and improved, or an open-source contribution you're proud of
Qualifications
Must Haves
- Strong Python and PyTorch skills
- depth in a relevant area
- the ability to run careful experiments
- Open to undergraduates, master's and PhD students, new graduates, and early-career researchers and engineers
Nice to Haves
- a project you took seriously, a paper you reproduced and improved, or an open-source contribution you're proud of
Benefits
- Remote within the U.S., with Bay Area hybrid preferred.
- Real-world data and modern GPU infrastructure.
- Work directly with the founders and technical team.