Principal Model Optimization Engineer

📋 Description

• Optimize ML models for GPU performance in training and inference.
• Profile performance to identify bottlenecks and improvements.
• Develop best practices and tooling for model optimization.
• Collaborate with data scientists and engineers to deploy models.
• Build tooling and visualizations to simplify ML at Roblox.

🎯 Requirements

• 6+ years of professional system design experience for scalable Roblox systems.
• Significant GPU debugging experience (GPU profiles, Xid errors).
• Proficient with CUDA, Triton, TensorRT to speed up models.
• LLM optimization: speculative decoding, batching, quantization.
• Passionate about squeezing GPU efficiency and ML performance.
• Advocate for generalization; build tools delivering model performance gains.
• Support internal partners (data scientists/ML engineers) needs.
• Bachelor's degree in CS, CE, Data Science, or similar.

🎁 Benefits

• Office roles: onsite Tue-Thu, with optional Mon/Fri.
• Equity compensation and benefits included.
• Equal opportunity employer; commitment to inclusion.