LLM Training Systems Engineer (Serverless, Multi-Tenant)

ByteDance; San Jose (CA); On-site; USD 128,000 - 256,000; Full time

Benefits offered by this job

Medical, dental, vision insurance

401(k) with company match

Paid parental leave

Short-term and long-term disability

Life insurance

Wellbeing benefits

10 paid holidays per year

10 paid sick days per year

17 days Paid Personal Time

Job summary

ByteDance's Volcano Ark team is seeking talented individuals to join the Data AML machine learning platform group. The role focuses on developing the Volcano Ark training system for post-training tasks such as SFT and RL, with emphasis on elastic, multi-tenant solutions across data centers and hardware.

Strong programming skills and training framework experience are required. Successful candidates will contribute to reinforcement learning interfaces and scalable training performance.

Qualifications

• Bachelor's or Master's in Computer Science or related field.
• Proficient in Python, Rust, or C++ with clean code and good abstraction.
• Experience in training framework design or training system optimization.
• Interest in tracking technical problems and optimization.

Responsibilities

• Develop Volcano Ark training system to enable post-training (SFT/RL) in a serverless setup.
• Design elastic training solutions for multi-tenant, multi-data-center environments to boost throughput and stability.
• Create reinforcement learning systems and user-friendly interfaces for training.

Skills

Python

Rust

C++

Framework design

Abstraction

Problem solving

Education

Bachelor's or Master's in Computer Science

Job description

ByteDance's Volcano Ark team is seeking talented individuals to join the Data AML machine learning platform group. The role focuses on developing the Volcano Ark training system for post-training tasks such as SFT and RL, with emphasis on elastic, multi-tenant solutions across data centers and hardware.

Strong programming skills and training framework experience are required. Successful candidates will contribute to reinforcement learning interfaces and scalable training performance.