---

---

Location

Brasil | Brazil

---

Job description

Location

Office Address

Project Description

Development of modern tools for working with AI / ML on the GPU. Implementation of algorithms for working with graphic processors. Working with GPU neural network calculations. Work with the AI/ML development team and the open-source community to analyze, develop, test, and deploy performance improvements for neural networks on GPUs with AMD ROCm, working on enabling ROCm HIPSOLVER/cuSOLVER in PyTorch

Responsibilities

Develop, support and optimize PyTorch project for GPU. Communication with developers, customers, and project managers. Create, document, and execute unit/component/functional tests. This might involve one or more of the following:
• Modifying PyTorch UT code to adapt it for ROCm
• Filing JIRAs on ROCm components that have missing features needed for the UTs

Skills

Must have

• Proficiency with PyTorch, С/C++ and Python
• Knowledge of GPU computing (HIP or CUDA or OpenCL)
• Experience in CMake, make/ninja build system, docker
• Software Version Control (Git)
• Knowledge in ML/NN
• GDB debugging

The successful candidate should possess the following traits:
• Proficiency in problem-solving and troubleshooting technical issues
• Time management, prioritization and multi-tasking skills
• Ability to handle multiple competing priorities in a fast-paced environment
• Personal initiative, commitment, perseverance and resilience
• Well-developed communication and teamwork skills
• Aspiration to engineering excellence

Nice to have

#J-18808-Ljbffr

---

Job tags

---

---

Salary