Distributed Training and Inference Engineer

Responsibilities

• Maintain, update, and optimize critical ML libraries and frameworks including JAX, PyTorch, CUDA, and ROCm
• Build and improve the entire ML software stack from ROCm/CUDA drivers to high-level JAX/PyTorch tooling
• Ensure model implementations are efficiently sharded and partitioned for large-scale distributed training and serving
• Integrate and validate modules for runtime correctness, memory efficiency, and scalability across multi-node clusters
• Conduct detailed profiling of compilation graphs and training workloads to eliminate bottlenecks
• Troubleshoot complex hardware–software interaction issues, including vLLM compilation failures and CUDA memory leaks

Requirements

• 5+ years of industry experience in ML systems or distributed training
• Bachelor’s or Master’s degree in Computer Science, Computer Engineering, or Electrical Engineering
• Strong programming experience in Python and C++
• Deep understanding of profiling tools such as Nsight, ROCm Profiler, or XLA profiler
• Expertise with partitioning configuration on modern ML frameworks like PyTorch and JAX
• Experience with multi-node distributed training systems and orchestration frameworks
• Hands-on experience with CUDA, ROCm, NCCL, or XLA

Preferred Qualifications

• Extensive experience with the XLA/JAX stack and compilation internals
• Familiarity with distributed serving frameworks like vLLM, TensorRT, or FasterTransformer
• Background in GPU kernel optimization or accelerator-aware model partitioning
• Understanding of low-level C++ building blocks used in ML frameworks

Benefits

• Medical, dental, and vision insurance
• 401k plan
• Daily lunch, snacks, and beverages
• Flexible time off
• Competitive salary and equity

About the Company

Sciforium is an AI infrastructure company developing next-generation multimodal AI models and a proprietary, high-efficiency serving platform.