Triton AI Compiler Architect for NPU Performance

Adecco; Warszawa; On-site; PLN 280,000 - 420,000; Full time

Benefits offered by this job

On-site in Warsaw

Job summary

Adecco is seeking a highly capable engineer to join our deep learning infrastructure team in Warsaw, Poland. You will own the Triton compiler and kernel optimization framework for NPUs, driving AI workloads with advanced performance engineering and compiler techniques.

You will implement state-of-the-art Triton kernels, optimize memory patterns, and ensure tight integration with PyTorch and other runtimes. This role combines AI compilation, NPU programming, and system performance, with mentoring

Qualifications

• Master’s or PhD in a relevant field and 5+ years in NPU/GPU programming or compiler development.
• Strong knowledge of accelerator architectures and performance bottlenecks.
• Proficiency in Triton, PTX, or LLVM IR for low-level optimization.

Responsibilities

• Lead design and development of the Triton compiler and performance framework for NPUs.
• Implement high-performance Triton kernels (Attention, MatMul, LayerNorm, Conv, Softmax).
• Optimize memory access, scheduling, and cache usage for scalable performance.
• Integrate Triton with PyTorch, XLA, and runtime stacks for end-to-end performance.
• Apply auto-tuning, kernel fusion, and operator scheduling to maximize throughput.
• Mentor teammates and establish processes for performance analysis and optimization.
• Stay current with MLIR, TVM, and related compiler tech to push boundaries.
• Perform workload fingerprinting for large models to guide optimizations.

Skills

NPU programming

Kernel optimization

Performance profiling

System design

Mentoring

Education

Master’s degree in Computer Architecture or related field

Tools

Triton

LLVM IR

PTX

TVM

MLIR

Cutlass

Job description

Adecco is seeking a highly capable engineer to join our deep learning infrastructure team in Warsaw, Poland. You will own the Triton compiler and kernel optimization framework for NPUs, driving AI workloads with advanced performance engineering and compiler techniques.

You will implement state-of-the-art Triton kernels, optimize memory patterns, and ensure tight integration with PyTorch and other runtimes. This role combines AI compilation, NPU programming, and system performance, with mentoring