Description
We are seeking a Solutions Architect to join our Top AI Labs team. The successful candidate will contribute to the development of open-source inference frameworks, develop and optimize KV cache offloading frameworks for LLM workloads, drive R&D on compute performance in distributed training, and study computational challenges in machine learning systems.
In this role, you will collaborate with the community to develop feature and operator development, performance optimization, and model support for open-source inference frameworks such as SGLang and vLLM. You will also develop and optimize KV cache offloading frameworks for LLM workloads, supporting multi-level cache offloading and reuse across CPU, SSD, and remote storage to improve inference efficiency.
Additionally, you will drive R&D on compute performance in distributed training, exploring methods and technologies for performance optimization. You will study computational challenges in machine learning systems, identify common needs and bottlenecks, and build example code, acceleration libraries, or frameworks accordingly.
To succeed in this role, you will need over 10 years of working experience in the technology industry, with a master's degree or above in computer science, mathematics, electrical engineering, automation, or related fields. You will also need strong interest in accelerated computing, parallel computing, and heterogeneous computing, with the motivation to explore these areas in depth.
Strong programming skills, with a good understanding of data structures and computer systems fundamentals, are essential. You will also need strong learning agility, adaptability, and the ability to analyze, define, and independently explore technical problems.
Familiarity with heterogeneous computing, distributed training, parallel computing, or other areas related to high-performance computing is desirable. Experience in performance analysis, performance modeling, or performance optimization, with contributions to open-source frameworks, is also a plus.