Kernel Performance Engineer — C++/GPU for AI Compute

Tenstorrent Unlisted/Referral Jobs; Warszawa; Hybrid; PLN 180,000 - 320,000; Full time

Job summary

Tenstorrent is building next-generation AI compute and seeks a strong C++ systems engineer to develop and optimize GPU‑style kernels. You will focus on performance, concurrency, and memory trade‑offs, working with cross‑functional teams to deliver accelerated workloads.

The role is hybrid in Warsaw or Gdansk, Poland, and offers opportunities to influence kernel design, tooling, and end‑to‑end throughput improvements in production systems.

Qualifications

• Strong C++ systems experience with performance‑critical code.
• Ability to reason about concurrency, synchronization, latency hiding, and memory trade‑offs.
• Data‑driven optimization using profiling/benchmarking results.

Responsibilities

• Design, implement, and optimize GPU‑style kernels such as matrix multiply and attention primitives.
• Own performance from bottlenecks to throughput improvements.
• Contribute to host‑side orchestration code and parallelization strategies.
• Develop micro‑benchmarks, regression tests, and tooling for correctness and performance.
• Collaborate with compiler, runtime, ML and hardware teams to productionize kernels.

Skills

C++ systems engineering

Concurrency & synchronization

Profiling & benchmarking

Debugging kernel-level issues

Performance optimization

Tools

Kernel development

GPU-style kernels

Job description

Tenstorrent is building next-generation AI compute and seeks a strong C++ systems engineer to develop and optimize GPU‑style kernels. You will focus on performance, concurrency, and memory trade‑offs, working with cross‑functional teams to deliver accelerated workloads.

The role is hybrid in Warsaw or Gdansk, Poland, and offers opportunities to influence kernel design, tooling, and end‑to‑end throughput improvements in production systems.