Intern - Research

Job Description

Research Internship (F/M/NB) - Efficient Neural Representation of Large-Scale Environments - La Forge

About Ubisoft

Ubisoft's 19,000 team members, working across more than 30 countries around the world, are bound by a common mission to enrich players' lives with original and memorable gaming experiences. Their commitment and talent have brought to life many acclaimed franchises such as Assassin's Creed, Far Cry, Watch Dogs, Just Dance, Rainbow Six, and many more to come. Ubisoft is an equal opportunity employer that believes diverse backgrounds and perspectives are key to creating worlds where both players and teams can thrive and express themselves. If you are excited about solving game-changing challenges, cutting edge technologies and pushing the boundaries of entertainment, we invite you to join our journey and help us create the unknown.

Ubisoft Bordeaux

Founded in 2017, Ubisoft Bordeaux works with passion on the biggest AAA titles to deliver the best gaming experiences. Today, the studio is composed of 400 talents from 20 different nationalities, working on licenses such as Assassin's Creed, Beyond Good & Evil 2 and a free to play game, BattleCore Arena. At the same time, the studio has set up a Tech branch which works on all Ubisoft's online services (named Online Services) as well as on the Anvil game engine. Ubisoft Bordeaux is also home to a R&D team, La Forge, which brings together engineers and researchers to work together on prototypes for game production, particularly around AI topics.

Job Description

A ray casting operation involves shooting a ray from a point into a 3D scene to detect intersections with objects. It is a fundamental task in game engines, forming the core of key systems such as rendering, collision detection, and even AI behavior. Given its widespread use across multiple subsystems, having a robust and efficient ray casting implementation is essential for real-time performance.

Ray casting is often accelerated with a bounding volume hierarchy data structure (BVH) [1, 2] to quickly find intersections, as it allows skipping empty space when tracing the ray. However, BVH traversal is an irregular algorithm, heavily influenced by the complexity and size of the scene, as well as the specific query (starting point and direction). This results in divergence in memory access and branch execution, making it less efficient on GPUs. Moreover, BVHs can have a significant memory footprint, especially when handling large, open worlds.

Neural methods have shown impressive potential for data compression and representation. More importantly neural network (NN) execution, especially fully connected ones, is considered a regular algorithm, relying on dense matrix multiplications with predictable memory access patterns, which are GPU-friendly. Recent research has explored replacing the BVH with neural networks, but most of these methods are focused on high-quality, dense objects [3, 4] or limit the network to output only visibility information [5].

The goal of this internship is to design an efficient neural representation capable of learning a large-scale scene and outputting high-dimensional information beyond simple visibility (e.g., distance, material semantics), providing a more comprehensive solution for ray casting in complex environments.

References :

[1] Meister D. et al. ''A survey on bounding volume hierarchies for ray tracing''. Computer Graphics Forum (2021).

[2] Meister D. et al. ''Performance comparison of bounding volume hierarchies for gpu ray tracing''. Journal of Computer Graphics Techniques (JCGT) (2022).

[3] Weier, P. et al. ''N-BVH: Neural ray queries with bounding volume hierarchies.'' ACM SIGGRAPH (2024).

[4] Fujieda, S. et al. ''Neural Intersection Function.'' arXiv preprint arXiv:2306.07191 (2023).

[5] Zhi Y. et al. ''Efficient Visibility Approximation for Game AI using Neural Omnidirectional Distance Fields.'' Proceedings of the ACM on Computer Graphics and Interactive Techniques (2024).

Qualifications
• Currently a second-year master's student or a third-year engineering student.
• Solid foundation in Machine Learning, linear algebra, and signal processing.
• Knowledge of computer graphics fundamentals, including Raytracing, is a plus.
• Proficiency in Python, and familiar with deep learning frameworks (e.g., PyTorch, TensorFlow).
• Familiarity with C++ is a plus.
• Proficient in English, both written and spoken, with the ability to clearly communicate technical concepts and collaborate effectively with an international team.

Jobcode: Reference SBJ-g3ez99-3-139-67-53-42 in your application.