Software Engineer (ML Infrastructure)

TL;DR

Software Engineer (ML Infrastructure): Building and scaling a serverless LLM post-training platform (Cortex Training) with an accent on distributed systems, GPU orchestration, and multi-node training. Focus on optimizing throughput, ensuring fault tolerance, and productionizing state-of-the-art research into enterprise-scale components.

Location: US-WA-Bellevue

Compensation: $160K – $230K

Company

Snowflake is a data cloud company powering the era of the agentic enterprise.

What you will do

• Design and build the full stack from public training APIs and SDKs to the GPU data plane.
• Scale distributed systems for serverless GPU compute, including multi-tenant scheduling and capacity-aware routing.
• Optimize end-to-end performance for training, inference, and RL loops to keep GPUs saturated.
• Partner with Snowflake Research to productionize state-of-the-art training and inference techniques into reliable components.

Requirements

• 5+ years building and shipping production ML systems.
• Strong foundation in distributed systems, designing fault-tolerant services on Kubernetes.
• Familiarity with GPU and LLM infrastructure (e.g., PyTorch, DeepSpeed/FSDP, Ray, CUDA/NCCL, vLLM).
• Demonstrated ability to harden complex systems for reliability and cost efficiency.
• BS in Computer Science or a related field.

Nice to have

• Hands-on LLM post-training or modeling experience.
• MS or PhD in Computer Science or a related field.