Senior ML Infra Architect — Scale & Optimize (Equity)

The Briefed; San Mateo (CA); On-site; USD 240,000 - 340,000; Full time

Benefits offered by this job

Equity compensation

Benefits for full-time employees

Job summary

Roblox in San Mateo is seeking a Machine Learning Infrastructure Engineer to build scalable, high-performance infrastructure powering ML models across Ads & Discovery. You will design systems for training and serving at scale, spanning transformers, large language models, and content-understanding components.

Collaborating with data scientists, core modelers, and platform teams, you will optimize GPU usage, implement efficient KV-cache strategies, and drive architecture and best practices for

Qualifications

• 5+ years designing and deploying large-scale ML systems in production.
• 3+ years leading ML infrastructure engineers.
• Strong communication and collaborative problem solving.
• BS, MS, or PhD in Computer Science, Engineering, or equivalent experience.

Responsibilities

• Co-design models and systems with modelers, data and AI infra engineers, and product teams.
• Investigate trade-offs end to end from data pipelines to low-latency inference.
• Design efficient KV-cache strategies, pruning, quantization, and GPU utilization.
• Work with FSDP, vLLM, SGLang, CUDA, and distributed frameworks for reliable performance.
• Lead planning and roadmap for production ML systems including training and inference.
• Own architecture and best practices for scalable, reliable ML infra.
• Collaborate with data scientists and platform teams to speed up model development.
• Stay current on ML infra industry trends and state-of-the-art practices.

Skills

Large-scale ML infra

FSDP

vLLM

SGLang

CUDA kernel development

Inference optimization

Education

BS/MS/PhD in CS/Engineering or equivalent

Job description

Roblox in San Mateo is seeking a Machine Learning Infrastructure Engineer to build scalable, high-performance infrastructure powering ML models across Ads & Discovery. You will design systems for training and serving at scale, spanning transformers, large language models, and content-understanding components.

Collaborating with data scientists, core modelers, and platform teams, you will optimize GPU usage, implement efficient KV-cache strategies, and drive architecture and best practices for