Research Scientist Intern, Spatial Reasoning & Localization (PhD)

Meta•Boston, MA

1d•$7,650 - $12,667

About The Position

Meta Reality Labs brings together researchers and engineers to invent the future of visual computing. Its teams develop technologies for capturing, creating, understanding, and interacting with immersive visual experiences across Meta products and AR/VR platforms. Research areas include artificial intelligence, computer vision, multimodal language models, maps, localization, scene text and landmark understanding, and spatial reasoning grounded in maps and floor plans. The Research Scientist Intern will advance a fast, lightweight vision-language-model-driven visual localizer that reasons over semantic and geometric evidence in images and structured map representations. The localizer estimates where an image was taken, including the place, floor, and camera position and heading. The intern will collaborate with researchers and engineers to develop models, representations, datasets, and evaluation methods, with opportunities to publish research and influence future products.

Requirements

Currently pursuing a PhD in computer science, electrical engineering, or a related field.
Experience in one or more areas of computer science, such as computer vision, machine learning, vision-language models, or visual localization.
A record of significant results demonstrated through grants, fellowships, patents, or co-authored publications at top-tier conferences such as CVPR, ICCV, ECCV, NeurIPS, ICLR, or ICML.
Experience with modern C++ and/or Python.
Experience with deep learning frameworks such as PyTorch.

Nice To Haves

Intent to return to a degree program after completing the internship.
Experience collaborating across groups and cultures.
Strong communication skills and ability to convey research ideas to a broad audience.
Research experience in visual localization, visual place recognition, image geolocalization, scene text recognition, or camera pose estimation.
Mathematical background in linear algebra, numerical optimization, and probabilistic estimation, and understanding of 3D geometry and camera models.
Experience training or fine-tuning vision-language models, including distillation, reinforcement learning, or efficient inference such as quantization or structured outputs.
Familiarity with map and place data, floor plans, or positioning signals such as GPS and Wi-Fi.
Relevant research or software engineering experience, including internships, work experience, coding competitions, or widely used open-source contributions.

Responsibilities

Research, develop, and prototype technologies in computer vision, multimodal machine learning, and visual localization, including compact vision-language models that extract structured spatial evidence from images and localize them against structured maps of indoor and outdoor environments.
Deliver research results that are publishable and have direct impact on Reality Labs or Meta’s Family of Apps.
Build datasets, metrics, and benchmarks for localization accuracy, latency, and cost, and compare results against frontier vision-language models.
Analyze the efficiency, accuracy, scalability, and stability of deployed technologies to identify opportunities for impact through novel research.
Collaborate with team members throughout projects, from research and prototyping through product deployment.