David Renaudie is Lead Data Scientist at Massive Entertainment - a Ubisoft Studio. With a Ph.D. in applied machine learning and a passion for solving real-world challenging problems with AI, he occupied in the past decade several positions in research & development and intellectual property in the industry. As a data scientist, he contributed to bring early machine learning prototypes to production-grade products, for several key projects.
In the gaming industry, he is leading a data science team that supports game productions with advanced data processing solutions, pursuing the goal of enhancing player experience and production tools. He constantly aims at pushing boundaries of what data science brings to the complex art of crafting AAA video games, and published in recognized conferences such as Game Developers Conference, IEEE Conference on Games or IEEE Big Data Conference. He is presently working on several projects related to main franchises of the studio, namely Tom Clancy's The Division 2, Avatar Frontiers of Pandora, and Star Wars Outlaws.
His team set themselves a challenge: could they predict the probable Xbox or PlayStation FPS of a scene from the PC workstation in real-time? If they could, content creators would be able to fix as they build and avoid the traditional build-test-fix loop.
The SnowCap plugin they developed sits inside Massive's SnowDrop game engine editor, predicting over 80% of problem areas. For best results, models are trained for each game build and console, but that's okay as the training data is easy(ish) enough for a production team to generate, and the benefits of early detection are clear.
David assured us that no LLMs were harmed in the making of SnowCap. A timely reminder that LLMs and GenAI are not the answer to every production problem.
The SnowCap plugin they developed sits inside Massive's SnowDrop game engine editor, predicting over 80% of problem areas. For best results, models are trained for each game build and console, but that's okay as the training data is easy(ish) enough for a production team to generate, and the benefits of early detection are clear.
David assured us that no LLMs were harmed in the making of SnowCap. A timely reminder that LLMs and GenAI are not the answer to every production problem.