AI Software Architect

Overview

Design, develop and optimize software features for AI frameworks, focusing on inference serving frameworks such as PyTorch, vLLM and SGLang, within Intel's Data Center Group AI software stack.

What you'll do

• Design and develop software features for AI frameworks, including hardware-agnostic and hardware-aware capabilities.
• Architect and analyze state-of-the-art features across frameworks to drive development throughout the software stack.
• Enhance and extend deep learning training and inference capabilities within Intel's AI frameworks software stack.
• Optimize software stack performance for deep learning workloads on GPUs and AI accelerators.
• Participate in discussions and development within open-source communities, contributing to upstream adoption and integration.
• Debug complex issues in multi-layered software systems and ensure seamless integration within large open-source frameworks.
• Conduct performance analysis using open-source and proprietary profilers to optimize code efficiency on host and accelerator platforms.
• Build technology demonstrators and proof-of-concepts for new features in the Intel AI Software Stack for current and upcoming GPUs.

What you'll need

• Bachelor's degree in Computer Science, Electrical and Computer Engineering, or related fields with 6+ years of industry experience; or Master's degree with 4+ years; or PhD with 2+ years.
• Proficiency in Python for developing complex software implementations.
• Intermediate expertise in advanced C++ programming (C++14/17), including parallel programming.
• Hands-on experience with frameworks such as PyTorch, vLLM, and SGLang.
• In-depth knowledge of Large Language Models (LLMs).
• Strong understanding of computer architecture and hardware-software optimization techniques.
• Proven ability to debug complex issues across multi-layered software systems.

Nice to have

• Practical experience with deep learning models for image and video generation.
• Expertise in developing and integrating kernels (e.g., GEMM, Convolution, Flash Attention) using frameworks like CUTLASS or Triton.
• Knowledge of compiler algorithms for heterogeneous systems and Fuser optimizations.
• Effective communication skills and experience collaborating with cross-geo teams.

Details

• Location: Bangalore, India.
• Work model: hybrid, splitting time between an assigned Intel site and off-site.
• Shift: Shift 1 (India).
• Job type: Experienced Hire.