About the job
We are seeking an experienced Research Scientist or Engineer to help define and build the next generation of AI infrastructure. In this role, you will work at the intersection of large-scale systems, AI, and emerging hardware to design infrastructure that enables reliable, efficient, and scalable AI workloads at ByteDance.
Responsibilities
AI Infrastructure Architecture
Design and evaluate scalable infrastructure architectures for large-scale ML workloads across compute, storage, and networking. Develop technical proposals and specifications that guide next-generation AI infrastructure systems.
Research & Technology Exploration
Track emerging trends in AI systems, distributed computing, and hardware acceleration. Conduct technical investigations and prototypes, and share insights through technical reports and presentations.
Performance & System Optimization
Analyze and optimize performance across the ML infrastructure stack—including scheduling, networking, storage, and training frameworks—through benchmarking, experimentation, and bottleneck analysis.
Cross-Team Technical Alignment
Work across research and engineering teams to translate AI workload requirements into scalable infrastructure solutions, providing architectural guidance and driving cross-team technical initiatives.
Qualifications
Minimum
- Master's degree or PhD in Computer Science, Electrical Engineering, or a related technical field.
- Strong proficiency in integrating AI tools into knowledge discovery and research workflows.
- 5 years of experience in distributed systems, infrastructure engineering, or ML systems. Experienced at evaluating trade-offs across hardware, software, and algorithms.
- Excellent communication skills to collaborate across teams.
Preferred
- Experience with large-scale model training and inference, including distributed training, KV cache–aware serving, GPU/accelerator optimization, and high-performance networking (e.g., RDMA, NCCL).
- Experience with heterogeneous AI compute systems, large-scale training clusters, HPC-style distributed workloads, and data pipelines for large model training and evaluation.
- Publications in systems and/or machine learning conferences (e.g., NeurIPS, OSDI, SOSP, ASPLOS, MLSys).
- Contributions to open-source projects.