Job Details

Overview

We are building a distributed LLM inference network that combines idle GPU capacity from around the world into a single cohesive plane of compute for running large-language models like DeepSeek and Llama 4. At any given moment, we have over 5,000 GPUs and hundreds of terabytes of VRAM connected to the network. We are a small, well-funded team working on difficult, high-impact problems at the intersection of AI and distributed systems. We primarily work in-person from our office in downtown San Francisco.

Responsibilities

• Design and implement optimization techniques to increase model throughput and reduce latency across our suite of models
• Deploy and maintain large language models at scale in production environments
• Deploy new models as they are released by frontier labs
• Implement techniques like quantization, speculative decoding, and KV cache reuse
• Contribute regularly to open source projects such as SGLang and vLLM
• Deep dive into underlying codebases of TensorRT, PyTorch, TensorRT-LLM, vLLM, SGLang, CUDA, and other libraries to debug ML performance issues
• Collaborate with the engineering team to bring new features and capabilities to our inference platform
• Develop robust and scalable infrastructure for AI model serving
• Create and maintain technical documentation for inference systems

Requirements

• 3+ years of experience writing high-performance, production-quality code
• Strong proficiency with Python and deep learning frameworks, particularly PyTorch
• Demonstrated experience with LLM inference optimization techniques
• Hands-on experience with SGLang and vLLM, with contributions to these projects strongly preferred
• Familiarity with Docker and Kubernetes for containerized deployments
• Experience with CUDA programming and GPU optimization
• Strong understanding of distributed systems and scalability challenges
• Proven track record of optimizing AI models for production environments

Nice to Have

• Familiarity with TensorRT and TensorRT-LLM
• Knowledge of vision models and multimodal AI systems
• Experience implementing techniques like quantization and speculative decoding
• Contributions to open source machine learning projects
• Experience with large-scale distributed computing

Compensation

We offer competitive compensation, equity in a high-growth startup, and comprehensive benefits. The base salary range for this role is $180,000 - $250,000, plus competitive equity and benefits including:

• Full healthcare coverage
• Quarterly offsites
• Flexible PTO

Skills: pytorch, gpu optimization, deep learning frameworks, sglang, vllm, cuda programming, machine learning, python, llm