Principal Model Optimization Engineer (ML)

TL;DR

Principal Model Optimization Engineer (ML): Optimizing machine learning models for performance on GPU architectures with an accent on training and inference workflows. Focus on conducting low-level performance profiling, reducing latency, and building scalable tooling for ML platforms.

Location: Must be based in San Mateo, CA (Hybrid: Onsite Tue-Thu)

Company

Roblox is a global platform for immersive 3D digital experiences, connecting millions of users through community-created content.

What you will do

• Optimize machine learning models for performance on GPU architectures, focusing on training and inference.
• Conduct low-level performance profiling to identify and resolve bottlenecks in ML pipelines.
• Develop best practices and tooling for model optimization and deployment.
• Collaborate with data scientists and software engineers to integrate optimized models into production.
• Build interfaces and visualizations to enhance the usability of the ML platform.

Requirements

• 6+ years of professional experience in system design and performance engineering.
• Significant experience debugging GPUs, including reading profiles and Xid errors.
• Proficiency in advanced frameworks such as CUDA, Triton, and TensorRT.
• Experience with LLM optimization techniques like speculative decoding and quantization.
• Bachelor's degree in Computer Science, Computer Engineering, Data Science, or a related field.

Culture & Benefits

• Equity compensation for all full-time employees.
• Comprehensive benefits package.
• Hybrid work environment with onsite presence required Tuesday through Thursday.
• Opportunity to solve unique technical challenges at massive scale.