3D Computer Vision Research Engineer, Geospatial Visual Localization

niantic-spatial; United Kingdom; Hybrid; GBP 70.000 - 120.000; Vollzeit

Zusammenfassung

Niantic Spatial, Inc. is seeking a Research Engineer to advance Geospatial Visual Localization technologies.

You will work with researchers in 3D vision and ML to publish at top venues and train large multi-modal models on extensive datasets, including our own 30B pose image collection, moving to production use for diverse global environments. You will work with devices from phones to drones and pursue robust, scalable localization under GPS-denied conditions while coordinating with product and

Qualifikationen

• Must have strong grounding in 3D computer vision, deep learning, robotics or related field.
• Experience with multi-view geometry and pose estimation, SfM, SLAM, georeferencing.
• Knowledge of geospatial data and coordinate reference systems.
• Excellent Python skills and ability to scale DL models using PyTorch.

Aufgaben

• Design, implement and evaluate novel visual localization algorithms combining 3D vision and deep learning.
• Train large multi-modal models and build benchmarks for accuracy and robustness.
• Develop robust solutions for noisy GPS, changing appearance and large extents.
• Collaborate with researchers and engineers to deploy production-scale solutions.
• Communicate results through technical writing, reviews and presentations.

Kenntnisse

3D computer vision

Deep learning

Robotics

Photogrammetry

Geospatial data

Python

Communication

Research collaboration

publishing/technical writing

Ausbildung

PhD in CV / ML

Tools

PyTorch

NumPy

SciPy

Jobbeschreibung

About Niantic Spatial

AI has learned from almost everything online, but it still struggles with the physical world, where 99.5% of economic activity happens. It doesn't know where things are, how spaces fit together, or what changes over time.

That's one of the biggest unsolved problems in AI, and it's the one we're solving at Niantic Spatial. We build spatial AI models and services so people and machines can reason, learn and act safely in the physical world. We turn camera imagery, video, sensor data and archives into accurate, AI-readable representations of real places that robots can train in and machines can use to know exactly where they are.

In physical AI, mistakes have real consequences. That's why we build AI grounded in measured reality, not guesswork, so the people and machines relying on it can trust it to get things right.

Our team built some of the world's most widely used maps. Now we're building something bigger: a living, measured record of the real world that machines can query, learn from and use to understand how places change over time. If you want to help AI understand the world we actually live in, come build with us.

About the Role

We are looking for a Research Engineer to help develop our Geospatial Visual Localization technologies, which allows people and robots to accurately localize themselves anywhere on the planet.

This is an end-to-end applied research role. You will develop new approaches to open problems alongside our researchers in 3D vision and machine learning, and publish your work at leading conferences. You will help build and train large models on large-scale multi-modal datasets, including our proprietary collection of 30 billion posed images and ship them into production, to be used by people and robots from all over the world.

Our team works with heterogeneous captures from phones and other handheld devices, 360 cameras, vehicles, and drones, as well as aerial and satellite imagery, LiDAR, and external geographic data. Your focus will be to improve accuracy and robustness across urban and rural locations, cameras, seasons, weather, and capture quality. The system must remain reliable with noisy GPS, in GPS-denied environments, and with outdated reference maps. Each pose estimate comes with an uncertainty that our customers rely on, and keeping it well calibrated across these conditions is part of the work.

What You'll Do

• Design, implement, and evaluate novel visual localization algorithms, combining 3D vision and deep learning.


• Train large models on multi-modal data at scale, and build the benchmarks that measure their accuracy and robustness.


• Develop robust solutions for difficult real-world conditions: noisy or denied GPS, changing appearance, varying capture quality, large geographic extents, and imperfect reference maps.


• Work side-by-side with leading researchers in visual localization, 3D computer vision, and mapping, as well as engineering and product teams, to identify the highest-impact problems and turn research into end-to-end solutions that operate at production scale.


• Communicate results clearly through technical writing, design reviews, internal presentations, and publications.

What You'll Bring

• Strong grounding in 3D computer vision, deep learning, robotics, photogrammetry, or a closely related field.


• Demonstrated experience with multi-view geometry and related problems, such as camera pose estimation, monocular depth estimation, SfM, SLAM, georeferencing, point-cloud registration, or sensor fusion.


• Knowledge of transformer models for 3D vision, such as vision foundation models, feed-forward 3D reconstruction, or multi-view and multi-modal attention.


• Knowledge of geospatial data and coordinate reference systems.


• Ability to formulate ambiguous real-world problems, choose appropriate mathematical or learning-based tools, and evaluate solutions rigorously.


• Strong Python skills and experience training deep learning models at scale, with libraries such as PyTorch, NumPy, and SciPy.


• Strong software engineering skills, with the ability to turn research prototypes into reliable, scalable, and maintainable systems.


• Experience taking an algorithm from research idea and experimental prototype through evaluation, benchmarking, and deployment on real-world data.


• Clear written and verbal communication and the ability to collaborate across research and engineering disciplines.

Nice to Have

• A PhD in computer vision, robotics, machine learning, photogrammetry, geodesy, remote sensing, or a related field, or equivalent research or industry experience. A PhD is not required.


• A strong publication record or substantial active involvement in the research community, especially at venues such as CVPR, ICCV, ECCV, NeurIPS, ICLR, SIGGRAPH, SIGGRAPH Asia, 3DV, ICRA, RSS, or related conferences and journals.


• Experience with aerial or satellite imagery, GNSS/GPS, IMUs, LiDAR data, geodesy, mapping systems, or large-scale spatial data.


• Evidence of technical lea