Research Kernel Engineer
BitDeer Technologies Group14 days ago
Singapore, SingaporeMid Level
Responsibilities
- Write and optimize CUDA and Triton kernels for methods developed by the research team.
- Perform performance attribution across the serving path, including roofline analysis and kernel-level bottleneck identification.
- Implement published methods whose reference implementations are too slow for production use.
- Optimize inference-critical operations such as attention, GEMM, normalization, sampling, and KV-cache management.
- Build measurement and benchmarking practices that support the broader research team.
- Collaborate closely with the AI Cloud platform team that owns the serving stack.
Requirements
- Bachelor's, Master's, or PhD in Computer Science, Electrical Engineering, or a related field.
- Hands-on experience with GPU programming, high-performance computing, or ML systems.
- Proficiency in CUDA and/or Triton, with concrete examples of writing or optimizing kernels.
- Strong Python and C++ skills.
- Working knowledge of GPU architecture and memory hierarchy, including occupancy, memory coalescing, tensor cores, and warp-level primitives.
- Experience using profiling tools such as Nsight Compute, Nsight Systems, or equivalent.
- Hands-on experience optimizing inference-critical paths such as attention, GEMM, normalization, sampling, or KV-cache management.
- Ability to implement methods without an existing reference implementation using papers, notebooks, or informal specifications.
- Ability to define appropriate measurements before optimizing and interpret performance results rigorously.
- Experience with vLLM, SGLang, or TensorRT-LLM, including custom kernels or extensions, is highly preferred.
- Compiler or IR-level experience with MLIR, TVM, or TorchInductor is highly preferred.
- Experience with distributed serving, including tensor or pipeline parallelism, is highly preferred.
- Publications at top-tier systems venues or substantial open-source contributions to inference or GPU computing projects are welcome.
- Strong interest in AI infrastructure, hardware performance optimization, ownership, and disciplined engineering.
Benefits
- Inclusive and respectful workplace that values diverse perspectives.
- Opportunity to work with industrial pioneers and contribute to new projects and systems.
- Personal accountability, autonomy, fast growth, training, mentoring, and other developmental opportunities.
- Attractive welfare benefits.
- Opportunity to contribute directly to the future of the digital asset industry.
Categories