Research Scientist Intern, TPU(PhD)

The PyTorch TPU team focuses on enabling efficient and accurate inference of 1P models directly on TPU. The work spans performance modeling, sharding and parallelism design, kernel authoring and automating with agents. Our internships are twelve (12) to twenty four (24) weeks long, and we have various start dates throughout the year.

Responsibilities

• Develop, evaluate methodologies and systems for running LLM/Multi-modality models cost-effectively on TPU systems.
• Analyze and extract insights from traffic pattern to derive computational intrinsics that can be leveraged into actions in system and kernel optimization.

Minimum Qualifications

• Currently has, or is in the process of obtaining, a PhD in computer science, computer engineering, electrical engineering, or a related field
• Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment
• Proficient in Python
• Experience with system and kernel optimization

Preferred Qualifications

• Experience with GPU/TPU/ASICs
• Experience with Triton, Cuda, Cutedsl, Pallas
• Familiarity with performance modeling