Senior Kernel Development and Optimization Engineer

About the role

Tenstorrent builds next-generation AI compute and has developed a high-performance RISC-V CPU from scratch. Its Kernel Development and Optimization team creates the performance-critical kernels that unlock the full capability of its hardware across ML and HPC workloads. This role is hybrid, based out of Belgrade, Serbia, and open to candidates at various experience levels; assessment during interviews determines the appropriate level.

Key responsibilities

• Design, implement and optimize GPU-style kernels such as matrix multiplication, attention primitives and data-movement operations.
• Take clear ownership of performance, from identifying bottlenecks to delivering measurable throughput improvements.
• Contribute to host-side orchestration code and parallelization strategies.
• Develop micro-benchmarks, regression tests and tooling to ensure correctness and sustained performance gains.
• Work closely with compiler, runtime, ML and hardware teams to integrate kernels into production systems.

Required profile

• A strong C++ systems engineer with experience writing performance-critical or low-level software.
• Comfortable reasoning about concurrency, synchronization, latency hiding and compute versus memory trade-offs.
• Data-driven, using profiling and benchmarking results to guide optimization decisions.
• Effective at debugging complex runtime or kernel-level issues in large codebases.
• Structured thinker who turns ambiguous performance problems into measurable experiments.

What you will learn

• The execution model, memory architecture and performance characteristics of Tenstorrent AI hardware.
• How to write and optimize accelerator kernels outside traditional CUDA-first ecosystems.
• Practical AI-assisted and agentic workflows for kernel generation, debugging and optimization.
• How low-level kernels, compilers, runtime systems and hardware co-evolve in modern AI platforms.