AI Libraries Developer

International Business Machines Corporation; Bengaluru; On-site; INR 2,200,000 - 3,200,000; Full time

Job summary

International Business Machines Corporation (IBM) in Bengaluru seeks an experienced AI Library Developer to advance AI runtime frameworks on IBM POWER architecture, with CPU kernel development and optimization.

You will collaborate with internal teams and the open-source community to upstream CPU-accelerated features, tune kernels, and analyze performance across Linux environments in a globally distributed team.

Qualifications

• Bachelor's degree in a relevant field.
• Strong knowledge of Computer Architecture, OS, Compilers, and Runtimes.
• 3+ years of C/C++ and Python development.
• 1+ year experience in assembly language and low-level debugging.
• 1+ year experience in AI workload optimization.
• Experience with open-source development tools such as Git, JIRA, or Bugzilla.
• Hands-on experience with AI libraries: OpenBLAS, ONNXRuntime, PyTorch, TensorFlow, Vector DBs.
• Solid foundation in Linear Algebra, Eigenvalues, and Numerical Computing.

Responsibilities

• Collaborate with internal teams and the open-source community to define, design, develop, and verify on Power-specific features in AI libraries such as OpenBLAS, ONNXRuntime, PyTorch, vLLM, Ollama and Vector Databases, focusing on CPU kernel development and optimization.
• Ensure CPU-accelerated features are upstreamed and integrated into mainstream Open-source distributions.
• Develop and tune CPU kernels to accelerate AI workloads.
• Perform performance analysis and tuning using POWER instruction profiling tools.
• Resolve issues and defects including problem recreation, debugging, proposing fixes, validating them, and upstreaming solutions with the community.
• Support customer-critical situations, lead proof-of-concept implementations, and provide technical guidance for pre-sales and post-sales engagements.

Skills

C/C++

Python

Assembly language

AI optimization

Open-source development

Education

Bachelor's Degree

Tools

Git

JIRA

Bugzilla

Job description

Your Role and Responsibilities:

This position is for an experienced AI Library Developer with strong expertise in AI runtime frameworks and libraries development and optimization. You will be part of the Power System Performance team, focusing on enhancing AI frameworks and libraries to leverage the full capabilities of IBM POWER architecture. Experience tuning for other architectures such as x86 or ARM is also highly relevant.

• Collaborate with internal teams and the open-source community to define, design, develop, and verify on Power-specific features in AI libraries such as OpenBLAS, ONNXRuntime, PyTorch, vLLM,Ollama and Vector Databases, with a focus on CPU kernel development and optimization.
• Ensure CPU-accelerated features are upstreamed and integrated into mainstream OPensource distributions.
• Develop and tune CPU kernels to accelerate AI workloads.
• Perform performance analysis and tuning using POWER instruction profiling tools .
• Resolve issues and defects including problem recreation, debugging, proposing fixes, validating them, and working with the community to upstream solutions.
• Support customer-critical situations, lead proof-of-concept implementations, and provide technical guidance for pre-sales and post-sales engagements.

This role requires deep understanding of CPU architecture, Linux-based OS environments, and AI workload optimization. Familiarity with software build systems, bring-up, testing, and performance analysis tools is desirable. You will work in a globally distributed team environment.

Required Education:

Bachelor's Degree

Required Technical and Professional Expertise:

• Strong understanding of Computer Architecture, Operating Systems, Compilers, and Runtimes.
• 3+ years of experience in C/C++ and Python development.
• 1+ year experience in assembly language and low-level debugging.
• 1+ year experience in AI workload optimization.
• Experience with open-source development tools such as Git, JIRA, or Bugzilla.
• Hands-on experience with AI libraries: OpenBLAS, ONNXRuntime, PyTorch, TensorFlow, Vector DBs.
• Solid foundation in Linear Algebra, Eigenvalues, and Numerical Computing

Preferred Technical and Professional Experience:

• Experience in low-level programming involving OS interaction, memory management, and threading.\
• Experience in GPU programming (CUDA/OpenCL) and AI workload optimization.
• Debugging, tracing, and profiling code for performance and functional issues.
• Familiarity with IBM AIX, Linux on POWER, or Z systems.
• Contributions to open-source projects or active participation in community mailing lists.
• Development experience with core libraries such as Glibc, libm, or PMDK.
• Exposure to compiler technologies (LLVM, MLIR) and graph optimization.

Years of Experience: 5 - 12