Senior LLM Inference Optimization Engineer at Intel

Compensation

$195,200 – $361,200/yr

Setup: Hybrid
Location: California · Phoenix, Arizona · +2
Type: Full-time
Level: senior

Description

Intel is seeking an experienced software engineer specializing in optimizing large language model inference engines for local and edge hardware. The role focuses on profiling and improving llama.cpp and vLLM performance across GPU/iGPU and CPU environments, tuning KV cache, batching, scheduling, quantization, startup, model loading, and lifecycle management, while contributing patches to open-source engines. The position requires a BS/MS in a related STEM field, 8+ years of software development, strong C++ or Python, systems-level Linux development, low-level debugging, and demonstrated CPU/GPU performance optimization experience. It is hybrid and based in Santa Clara, California, with additional locations in Phoenix, Arizona; Folsom, California; and Hillsboro, Oregon.

Required Skills

C++
Python
llama.cpp
vLLM
KV cache
Quantization
Performance profiling
Linux
Low-level debugging
Vulkan
GPU programming
Benchmarking
Open-source contributions

Additional Information

Minimum Education: Bachelor's or master's degree in computer science, electrical engineering, mathematics, or a related STEM field
Required Previous Experiences: 8 or more years of software development experience
Experience with large language model inference, including attention, KV cache, and decoding
Experience profiling and optimizing CPU or GPU performance problems with demonstrated speedups
Experience with Linux, build systems, and low-level debugging
Preferred Qualifications: Hands-on experience with llama.cpp, vLLM, ggml, or similar inference engines
GPU or accelerator programming with Vulkan, CUDA, SYCL, or Metal, or experience with SIMD and CPU kernels
Familiarity with quantization formats and quality trade-offs
Open-source contributions to inference engines