3D machine learning
S-au achitat serviciile după ce au fost prestate
We are looking for a ML Engineer with expertise across multiple 3D AI networks. This position will focus on evaluating, implementing and innovating upon various 3D networks for asset and world creation within virtual worlds, namely video games.
This person should be a skilled coder, capable of building on top of SOTA AI models and integrating various models together to create higher quality results. In addition, given our work in the creative industries having aesthetic and geometry processing skills are a plus.
The nature of these types of jobs tends to result in a large amount of applications, and therefore to evaluate candidates we are asking you to perform a (we hope relatively simple) task aimed at showing some basic technical proficiency and more importantly your thinking.
Please note we will follow up with all candidates who successfully complete the task but will not follow up with candidates who do not do the task or are unsuccessful in completing it.
Please create three repos in google colab each based on a different type of SOTA 3D AI model. Using each repo generate a single object - a medieval house - and use / create / decide on a scoring metric to evaluate which of the three repos generated the 'best' medieval house. Summarise your work in a report and send a link to your report + links to your 3 colab repos as the completion of the task.
Please note, there are elements of this task intentionally vague so as to allow your creativity in completing it, e.g. which AI networks you choose to use, what metric(s) you choose to score, and even to some extent what you consider as a generated result for 'medieval house'.
Based on this report and a review of your colab models we will select candidates to interview.
Please note - this task is not the project itself, it is simply a way to distinguish and select candidates that have the skills necessary to work on the main task which will be building atop SOTA 3D AI models to create a new model and evaluation criteria for generated mesh results, which can be used in the gaming industry.
ID Proiect: #37489769