GPU Kernel Architect for High-Performance AI Inference

Bitdeer (NASDAQ: BTDR); Austin (TX); On-site; USD 120,000 - 180,000; Full time

Job summary

Bitdeer AI Lab seeks a research engineer to accelerate GPU-based AI workloads. You will implement and optimize CUDA and Triton kernels, analyze performance, and adapt cutting-edge inference methods for production use. Collaboration with AI Cloud platform teams is expected, driving impactful research into real systems.

The role emphasizes hands-on kernel development, strong CUDA/Triton skills, and a background in HPC or ML systems, with ownership and rigorous engineering discipline.

Qualifications

• Hands-on experience in GPU programming, HPC, or ML systems.
• Proficiency in CUDA and/or Triton with concrete kernel examples.
• Knowledge of GPU architecture and memory hierarchy; profiling habits to back it up.

Responsibilities

• Write and optimize CUDA / Triton kernels for team-developed methods.
• Perform performance attribution across serving path; roofline analysis.
• Make reference-methods production-viable when needed; build measurement basis for team.

Skills

GPU programming

CUDA

Triton

Python

C++

High-performance computing

ML systems

Profiling

Education

Bachelor's/Master's/PhD in Computer Science or Electrical Engineering

Tools

Nsight Compute

Nsight Systems

MLIR

TVM

TorchInductor

TensorRT-LLM

Job description

Bitdeer AI Lab seeks a research engineer to accelerate GPU-based AI workloads. You will implement and optimize CUDA and Triton kernels, analyze performance, and adapt cutting-edge inference methods for production use. Collaboration with AI Cloud platform teams is expected, driving impactful research into real systems.

The role emphasizes hands-on kernel development, strong CUDA/Triton skills, and a background in HPC or ML systems, with ownership and rigorous engineering discipline.