Senior AI Compiler Architect for Efficient Deep Learning

Ampere; Santa Clara (CA); On-site; USD 195,000 - 292,000; Full time

Benefits offered by this job

Health insurance

401K plan

Unlimited flextime

Job summary

Ampere, a semiconductor design company, is seeking a senior developer to optimize deep learning graphs and unlock the full potential of its AI accelerator. You will work across the SW/HW stack from inference serving to compiler kernels.

Requires a strong foundation in CS, solid Python/C++ skills, and experience with graph algorithms. Knowledge of LLVM/MLIR is a plus and success in competitive programming is advantageous.

Qualifications

• Bachelor's degree in Computer Science, Mathematics or related field with 8+ years of experience, or Master's with 6+ years.
• Strong CS fundamentals: algorithms, data structures, systems.
• Proficiency in Python and C/C++.
• IOI/ICPC medal or equivalent is a big plus.

Responsibilities

• Optimize deep learning computational graphs for performance, throughput, and latency on Ampere's accelerator hardware.
• Enable popular models and frameworks (PyTorch, Llama.cpp) and serving platforms (vLLM, SGLang).
• Identify and implement graph-level optimizations: op fusion, pattern recognition, redundancy elimination.
• Collaborate on HW/SW co-design to push computational efficiency.
• Work with cross-functional teams to integrate AI solutions into Ampere's AI hardware platforms.

Skills

Python

C/C++

Graph algorithms

Systems programming

Competitive programming

Education

Bachelor's degree in Computer Science or related field

Master's degree in Computer Science or related field

Tools

LLVM compiler framework

MLIR dialects

Job description

Ampere, a semiconductor design company, is seeking a senior developer to optimize deep learning graphs and unlock the full potential of its AI accelerator. You will work across the SW/HW stack from inference serving to compiler kernels.

Requires a strong foundation in CS, solid Python/C++ skills, and experience with graph algorithms. Knowledge of LLVM/MLIR is a plus and success in competitive programming is advantageous.