Artificial Intelligence and neural networks have only recently begun to be used in many areas of computer graphics, including photogrammetry, segmentation, and differential rendering. Since our company's goal is to achieve better quality with all possible tools, we have also begun to explore the use of deep neural networks in tasks related to the recovery of geometry, materials, and other important data for procedural scene generation, speeding up artists' work, and optimizing performance.
You will:
We require a specialist(s) who will work to define a roadmap for applied research into the application of AI to multiple problems, actively communicate with fellow engineers and artists, and propose, as well as implement, solutions sharpened to meet bottom-line business goals.
You have:
- Systems knowledge of Data Science, Deep Neural Networks, mathematical analysis, linear algebra, mathematical statistics
-
An understanding of the modern scientific picture of the world on the application of DNNs to popular problems such as segmentation, classification, especially for working with multiple images (ldr\hdr).
-
Ability to read scientific articles and select only the most efficient algorithms, taking into account all technical and business constraints of the applied research.
-
Ability to test hypotheses quickly and cheaply, to search various kinds of neural network architectures, hyperparameters, datasets
-
Experience in automating experiments, and logging them in detail
-
Experience in writing small reports and mini-presentations on the results of experiments.
-
Ability to communicate with specialists from many different disciplines (graphics, physics, artists).
-
Experience working with the most popular technical frameworks, languages, and tools for DNN: Pytorch\Tensorflow\Numpy\Sql and so on
A plus will be:
- Experience drafting an applied research plan and implementing it
-
Basic knowledge of computer architectures, theory of algorithms and data structures, as well as optimization and speed of code execution on modern CPUs and GPUs
-
Writing CUDA and/ or GLSL/ GLSL code under DirectX/Vulkan for GPU computing
-
Experience in bringing a finished model with all the pipelining (data transformation and compression, etc.) to execution in a working invironment to maximize performance. Benchmarking the performance of neural network models. Optimization of models through simplification/architecture modification and quantization of weights, layer merging, etc. Writing custom layers under TensorRT using tensor computation.
-
Understanding the basics of rendering and computer graphics. PBR, Rasterization, Materials, Geometry. Ray-Tracing
-
Up-to-date knowledge of the scientific picture of differential rendering and neuronics for comp. graphics. NeRFs, Mitsuba 2, Neural Radiance Cache, etc. Metrics for image comparison
.
