Postdoc: Open-Source OCCL for Distributed AI

1000scholars; Zürich; On-site; CHF 46,000 - 59,000; Full time

Benefits offered by this job

25 days annual leave

Healthcare insurance

NS mark certification

Flexible hybrid work arrangement

Job summary

The Singapore-ETH Centre invites applications for a Postdoctoral Researcher (OCCL) to co-lead the development of an open-source accelerator communication library and advance distributed AI workloads.

The role involves designing primitives, optimizing communication patterns, collaborating with partners in Singapore and Zurich, publishing results, and supervising MSc/BSc students. Contract duration: 2 years; strong English required.

Qualifications

• PhD in Computer Science or related field.
• Strong background in parallel computing, compilers and HPC systems.
• Excellent knowledge of C and C++.
• Experience with MPI, NCCL, UCX and high-speed networks.
• Proficient English communication; able to work in a distributed team.

Responsibilities

• Co-lead the design and development of an open-source accelerator communication library (OCCL) supporting point-to-point and collective communication for distributed AI workloads.
• Design and implement communication primitives and algorithms optimised for ML training and inference workloads.
• Investigate and optimise communication patterns for large-scale DL workloads and overlap with computation.
• Ensure OCCL is portable across hardware platforms and network substrates (UCX, NVLINK, InfiniBand, UltraEthernet).
• Evaluate and optimise the library on GPU clusters and large-scale systems; publish results and collaborate with partners in Singapore and Zurich.
• Contribute to supervising MSc and BSc students and communicate with supervisors and external partners.

Skills

C++

C

Parallel computing

High-performance computing

Performance optimization

Education

PhD in Computer Science or related field

Tools

MPI

NCCL

UCX

NVLINK

InfiniBand

UltraEthernet

Job description

The Singapore-ETH Centre invites applications for a Postdoctoral Researcher (OCCL) to co-lead the development of an open-source accelerator communication library and advance distributed AI workloads.

The role involves designing primitives, optimizing communication patterns, collaborating with partners in Singapore and Zurich, publishing results, and supervising MSc/BSc students. Contract duration: 2 years; strong English required.