AI System Performance Engineer

Overview

Build analytical models of AI training and inference systems to project end-to-end performance, locate architectural bottlenecks, and recommend optimal system configurations for Intel architecture.

What you'll do

• Enable new AI workloads in analytical simulators, and model emerging multi-GPU deployment, parallelization, and communication strategies as simulator capabilities.
• Build analytical models of newly proposed GPU architectures, including GPU core, memory hierarchy, and training/inference deployment scenarios, and produce performance estimates that expose architectural bottlenecks and drive improvement recommendations.
• Integrate and modify open-source simulators or analytical models for Intel-specific architecture.
• Perform what-if analysis by sweeping core count, memory/cache capacity and bandwidth, fabric bandwidth, GPU interconnect, and deployment options; deliver optimization and optimal architecture recommendations.

What you'll need

• BTech or Master's degree in Electrical Engineering, Computer Engineering, Computer Science, or a related field.
• College graduate or 1+ years of experience in AI workload and GPU architecture modelling/simulation or a similar field.
• Strong proficiency in Python, with the ability to build modular, maintainable analytical models and simulation frameworks.
• In-depth understanding of computer architecture: GPU/CPU, ALU/SM core, cache, and different memory types and hierarchy.
• In-depth understanding of LLM AI workloads (Transformer, MoE, Attention, MLA - LLaMA, DeepSeek, Qwen, and robotic workloads) and kernel internals.

Nice to have

• Understanding of distributed training/inference and parallelism strategies (data, pipeline, tensor, and expert) and their communication patterns when scaling across multi-GPU hardware.
• Understanding of LLM inference and training mechanics: prefill vs. decode, KV cache, batching policies, and quantization.
• Exposure to TRT-LLM, vLLM, or SGLang.
• Familiarity with Git/GitHub, CI/CD, and automated workflows.
• Understanding of power fundamentals and power/performance co-optimization.

Details

• Location: Bangalore, India.
• Work model: hybrid, splitting time between on-site at the assigned Intel site and off-site.
• Shift: Shift 1 (India).
• Job type: College Grad.