LPU Operator / Compiler / Performance Engineer

About the role

Campus Singapore is looking for an engineer to build LLM inference software for its Language Processing Unit (LPU). You will develop and optimize inference operators and kernels, and tune the compiler's lowering from tensor graphs to the LPU instruction set. Experience in operator development, compiler development or performance optimization is sufficient.

Key responsibilities

• Develop and optimize inference operators and kernels such as GEMM and attention, tuning tiling, data layouts and data movement, and validating correctness and performance across tensor shapes and precisions.
• Implement and optimize lowering from tensor graphs/IR to LPU ISA, including operator fusion, on-chip buffer allocation, instruction scheduling and code generation.
• Run microbenchmarks and end-to-end profiling to find compute, memory-access and synchronization bottlenecks, and validate first-token latency, per-token latency and throughput gains across input lengths and batch sizes.
• Co-design with architecture and runtime teams, evaluate trade-offs between compiler scheduling, operator implementations and hardware resource usage, and add key workloads to regression tests.

Required profile

• Recent or new PhD in CS, EE, Automation or a related field; exceptional final-year Master's students with strong research or engineering records are also welcome.
• Strong computer architecture fundamentals: parallelism, memory hierarchy and data dependencies.
• Able to break down problems independently, design reproducible experiments and communicate technical contributions clearly.

Required skills

• Proficiency in C/C++ and Python scripting; Linux-based development and debugging.
• Preferred: experience with LLVM, MLIR, TVM or Triton, or low-level programming with CUDA or other accelerators.
• Preferred: LLM inference optimization experience such as prefill/decode profiling, KV cache management, low-precision computation or cross-chip communication.

What we offer

• Mentorship from architecture, compiler and runtime experts.
• End-to-end ownership of a core LPU inference operator or compiler optimization, from prototype to regression validation.
• Access to state-of-the-art foundation models and LLM agents to accelerate R&D, generate tests and validate results.