Developer of Parallel Computing Systems on GPU/NPU for Autonomous Transport

We have set ourselves a large-scale goal: to be among the world's first companies to create autonomous vehicle control technology.

To accelerate the operation of the recognition and prediction system and, as a result, teach cars and delivery robots to act even faster in traffic situations, we use machine learning algorithms. The algorithms are becoming more complex, their number is growing, and our task is to increase their efficiency with limited resources.

If you are passionate about optimizing and parallelizing algorithms, and are ready to put every watt consumed by the equipment to good use, then you will definitely enjoy working with us!

What tasks await you

Improving algorithm efficiency You will optimize algorithms, taking into account the hardware features of platforms, automate algorithm testing and their adaptation to the platform.

Working with computing platforms You will explore modern platforms for executing parallel computations and select the optimal architecture for parallel execution of algorithms on a specific platform.

More about the backend at Yandex — in the channel Yandex for Backend

We expect that you

• Have worked with PyTorch (or TensorFlow), converted ML models to ONNX and ran them on TensorRT, ONNX Runtime, TFLite or any other framework
• Have used CUDA/OpenCL, optimized and parallelized computations on CUDA/OpenCL or other accelerator architectures
• Have a good knowledge of C++, Python
• Are proficient in classical data structures and algorithms
• Are familiar with linear algebra, probability theory
• Can explain your decisions and defend them against opponents
• Are able to work in a team

It will be a plus

• Have solved applied problems using machine learning or Deep Learning
• Have written custom plugins for ONNX, TensorRT or any other framework using C++, CUDA, OpenCL
• Are familiar with neural network compression, pruning, quantization methods
• Understand discrete mathematics and optimization theory
• Have a good knowledge of STL
• Have experience in industrial development not related to machine learning