QNT Partners Hiring! Full Time ML Research Platform Engineer (Distributed Training - HPC) in - Ricebowl

Location: Singapore, Hong Kong or Shanghai

About the role

We are looking for a platform engineer to build the infrastructure that powers our next-generation machine learning research. Think: large-scale experimentation, distributed training, and reproducibility.

This is not an applied ML role. You will not be fine-tuning LLMs or building agents. Instead, you will build the systems that enable researchers to train models at scale

What you will own

• Distributed training pipelines for GPU-accelerated workloads (PyTorch, JAX)
• Experiment management and model versioning
• Resource scheduling on on-premise HPC clusters and cloud (Slurm, Kubernetes)
• Observability and debugging for complex training jobs
• Data lineage and artifact tracking

Must haves (non-negotiable)

• 2+ years building large-scale distributed systems for research or data-intensive workloads
• Strong **Python** and experience writing high-performance, maintainable code
• Deep familiarity with modern ML frameworks (PyTorch, TensorFlow, or JAX)
• Hands-on experience with GPU-based training across multiple nodes
• Comfortable working directly with researchers or quantitative analysts

Strong pluses

• Experience with Slurm, PBS, or similar HPC schedulers
• Built or maintained experiment tracking and model versioning systems
• Worked in quantitative finance, simulation, or other performance-sensitive domains
• Experience with DeepSpeed, FSDP, or Megatron-LM
• Familiarity with NCCL, Ray, or Kubeflow

Not a fit if

• Your recent work is focused on LLM agents, RAG, chatbots, or inference serving
• You have not worked with distributed training across multiple GPUs
• You prefer building user-facing products over research infrastructure