Introduction

Are you eager to transform groundbreaking research into real-world solutions? Do you have a passion for developing cutting-edge AI systems that drive meaningful impact? Would you enjoy publishing in top-tier AI conferences and contributing to open-source advancements?

If so, we invite you to apply for our Research Scientist position at IBM. We are seeking highly motivated candidates with expertise in multimodal Large Language Models (LLMs) and a passion for hardware-software co-design. As part of our world-class research team, you will explore and develop novel hardware-algorithm co-design techniques to enhance AI model training and deployment.

Your work will contribute to SoC architecture and compiler stacks for next-generation hardware accelerators along with algorithmic breakthroughs for AI applications across multiple domains. You will collaborate with a diverse team of researchers specializing in algorithm optimizations, hardware architectures, graph compilers, AI frameworks, and applications.

Your role and responsibilities

The job responsibilities include working with a team of world-class researchers to help define and develop novel hardware architectures as well as software stacks, tools and methodologies that enable cutting edge research in Deep Learning related projects. The primary mission for the team is to lead the definition of the SoC architecture and compiler stack for new hardware accelerators for Deep Learning. Key responsibilities will include proposing new concepts in compilers to exercise the hardware architecture features of AI accelerators, implementing the compiler optimizations to generate programs and evaluating performance benefits using the software simulators across different use-cases, determining hardware/software co-techniques to improve performance. Further responsibilities include performing research and development of novel hardware/software solutions and evaluating their merits relative to state-of-the-art solutions, and demonstrating external eminence by publishing the outcomes of the research.

Required technical and professional expertise

Strong software engineering background, and proficiency in C, C++ and Python programming as well as experience with state-of-the-art software collaboration practices is essential.

Strong background in architecture and compilers with experience developing simulators and compilers.

Strong interest and passion to drive new research topics in hardware/software co-design for domain-specific System-on-Chip architectures.

Preferred technical and professional experience

Experience with the use and adaptation of Deep Learning frameworks including TensorFlow and PyTorch as well as CUDA for GPU programming is preferred, but not necessary

Strong background in deep-learning applications including large language models.