Careers

Join MarsCompute

Build faster, more efficient AI inference.

MarsCompute is an AI infrastructure startup focused on making large language model (LLM) inference faster, more efficient, and more affordable. We build production-grade AI inference systems and develop automated GPU kernel optimization technologies.

  • Singapore or United States
  • Full-time
  • Experienced engineers

Open positions

AI Inference / Product Engineer

Focus: Production LLM serving and AI infrastructure

What you'll do

  • Build and operate high-performance LLM inference services on GPUs.
  • Optimize inference throughput, latency, GPU utilization, and serving costs.
  • Develop reliable production systems, including deployment, monitoring, autoscaling, and fault recovery.
  • Work with inference frameworks such as vLLM, SGLang, and related technologies.
  • Bring new AI models into production quickly and reliably.

What we're looking for

  • Strong experience in Python, C++, Linux, and distributed systems.
  • Hands-on experience building and deploying production systems.
  • Experience with LLM serving, GPU computing, cloud infrastructure, or performance engineering.
  • Familiarity with Kubernetes, Docker, and CI/CD is a plus.
  • Ability to take ownership of engineering projects and deliver working products.

GPU Kernel / ML Systems Optimization Engineer

Focus: GPU performance and automated kernel optimization

What you'll do

  • Design and optimize high-performance GPU kernels for AI workloads.
  • Develop automated kernel optimization techniques using compiler technology and AI agents.
  • Optimize LLM inference operators, including attention, GEMM, MoE, and kernel fusion.
  • Analyze GPU performance bottlenecks across computation, memory, and communication.
  • Build benchmarking, profiling, correctness validation, and optimization pipelines.
  • Explore next-generation ML systems optimization for emerging models and GPU architectures.

What we're looking for

  • Strong programming skills in C++ and CUDA, Triton, or related GPU programming frameworks.
  • Experience with GPU architecture, parallel computing, compilers, or ML systems.
  • Hands-on experience with performance profiling and low-level optimization.
  • Familiarity with PyTorch, CUTLASS, CuTe, or ML compiler frameworks is a plus.
  • Experience in GPU kernel optimization, AI inference, or systems research is highly valued.

Why MarsCompute?

  • Real-world impact. Work on technologies deployed in production AI inference systems.
  • Deep technical challenges. Solve hard problems spanning GPU kernels, compilers, distributed systems, and AI.
  • Research meets engineering. Work on cutting-edge optimizations and turn new ideas into production-ready technology.
  • Ownership. Join an early-stage team where your engineering contributions directly shape our products.

How to apply

We value strong engineering skills, hands-on experience, and demonstrated technical achievements more than job titles.

We welcome candidates worldwide, including those currently based in China. We can assist qualified candidates with work visa applications and relocation, subject to applicable immigration requirements.

Please send us:

  • Your CV or LinkedIn profile
  • GitHub profile, technical projects, or selected publications (if available)
  • A short introduction to your engineering experience and preferred position

Contact: [email protected]