AI Infrastructure & ML Systems Intern

ByteDance; San Jose (CA); On-site; USD 97,000 - 138,000; Full time

Benefits offered by this job

Health insurance

Housing allowance

Paid holidays

Job summary

ByteDance Seed Infrastructure Internship in San Jose invites PhD candidates to contribute to distributed training, reinforcement learning frameworks, and high-performance inference for AI foundation models. You’ll explore data/model/pipeline parallelism, memory efficiency, and heterogeneous hardware compilation across GPUs and accelerators.

You will work on system-level performance analysis, profiling, and tooling to improve developer productivity and reliability.

Qualifications

• Pursuing a PhD in CS, EE, or related field.
• Strong programming skills in Python and/or C++.
• Solid understanding of systems, distributed computing, ML systems, or performance optimization.
• Experience with distributed training frameworks (e.g., PyTorch FSDP, Megatron-style parallelism).
• Experience with reinforcement learning training systems or GPU/compiler technologies.

Responsibilities

• Design and optimize large-scale distributed training systems (data/model/pipeline parallelism, memory efficiency, fault tolerance).
• Contribute to reinforcement learning training frameworks and large-scale post-training systems.
• Improve inference performance, latency, and throughput for foundation models.
• Develop compiler or runtime optimizations for heterogeneous hardware (GPU/accelerator).
• Work on system-level performance analysis, profiling, and bottleneck diagnosis.
• Build tooling and automation to improve developer productivity and system reliability.

Skills

Python

C++

Distributed systems

Machine learning systems

Performance optimization

CUDA

Triton

Education

PhD in Computer Science or Electrical Engineering

Tools

PyTorch FSDP

Megatron-style parallelism

CUDA

Triton

Job description

ByteDance Seed Infrastructure Internship in San Jose invites PhD candidates to contribute to distributed training, reinforcement learning frameworks, and high-performance inference for AI foundation models. You’ll explore data/model/pipeline parallelism, memory efficiency, and heterogeneous hardware compilation across GPUs and accelerators.

You will work on system-level performance analysis, profiling, and tooling to improve developer productivity and reliability.