Cloud Inference Engineer, Modular

• We're seeking engineers who are passionate about pushing the boundaries of distributed inference systems and enjoy working at the intersection of large-scale systems and machine learning
• We are looking for candidates based on their breadth and depth of experience in backend engineering, AI inference, and distributed systems development
• 5+ years of experience working in backend engineering
• Experience with kubernetes and operating your own services
• Ability to create durable, reusable software tools and libraries that are leveraged across teams and functions
• Experience in machine learning technologies and use cases
• Creativity and curiosity for solving complex problems, a team-oriented attitude that enables you to work well with others, and alignment with our culture
• Strongly identifies with our core company cultural values

Desirable

• In the Cloud Inference team, we are focused on building end to end distributed LLM inference deployments that are fully vertically integrated with the MAX stack
• Our goal is to make inference both the fastest and most scalable while also building an easiest platform for deploying and scaling models for enterprises and developers alike
• If this sounds exciting, we invite you to join our world-leading AI infrastructure team and help drive our industry forward!
• Build & ship a LLM focused inference platform using best in class inference techniques (disaggregated inference, multi-node deployment of large models, high performance networking, distributed kv-cache management, high throughput batch processing, etc)
• Push the envelope for operational excellence with request-to-kernel observability, multi-cloud deployments, clever autoscaling, cold-start optimizations, and more
• Collaborate with our kernels and genAI teams to achieve SOTA application performance by integrating SOTA kernel & serving optimizations with SOTA cluster optimizations
• Build helm charts, kubernetes operators, and more to make a create simple, effective, maintainable deployments