Senior AI Accelerator Software Architect (Framework & Perf) - Remote

Ampere Computing; Warszawa; Hybrid; PLN 417,000 - 625,000; Full time

Benefits offered by this job

Premium health care

Paid time off

Office amenities

Remote work support

Sports card

Job summary

Ampere Computing is seeking an AI Accelerator Software Senior Principal Engineer – Framework Integration to lead strategy for high-performance deep learning inference across Ampere accelerator platforms. You will set direction for enabling major ML frameworks and optimizing for throughput, latency, and efficient compute across data centers to edge.

This role requires deep architectural ownership, cross-team influence, and mentorship to raise the bar for quality, maintainability, and performance

Qualifications

• BS/MS/PhD with 12/8/5+ years experience respectively, in relevant field.
• Proven experience with integrating PyTorch, ONNX, and llama.cpp into accelerators.
• Linux accelerator runtime/driver experience is a plus.

Responsibilities

• Lead end-to-end framework integration for AI accelerators.
• Drive full-stack SW/HW acceleration from inference to kernel optimization.
• Improve model performance and correctness for PyTorch/ONNX/llama.cpp ecosystems.
• Provide HW/SW co-design directions to maximize efficiency and throughput.
• Build and evolve software components for AI accelerators.
• Collaborate across compiler/runtime/kernel and systems teams from cloud to edge.
• Mentor and raise standards for quality and performance.

Skills

Python

C/C++

Performance engineering

Linux experience

ML concepts

Framework integration

Education

BS in CS/CE/EE or related field

MS in CS/CE/EE

PhD in related field

Tools

PyTorch

ONNX

llama.cpp

vLLM

SGLang

Job description

Ampere Computing is seeking an AI Accelerator Software Senior Principal Engineer – Framework Integration to lead strategy for high-performance deep learning inference across Ampere accelerator platforms. You will set direction for enabling major ML frameworks and optimizing for throughput, latency, and efficient compute across data centers to edge.

This role requires deep architectural ownership, cross-team influence, and mentorship to raise the bar for quality, maintainability, and performance