Backend Engineer – ML Platform & Scheduling

ByteDance; Singapore; On-site; SGD 90,000 - 150,000; Full time

Job summary

ByteDance in Singapore is seeking a talented software engineer to design and develop resource scheduling systems for machine learning platforms, supporting training, evaluation, and inference workloads across NLP, CV, and Speech.

You will optimize the orchestration of GPUs, CPUs, and accelerators, and design workload scheduling across data centers and cloud environments to ensure efficient resource utilization and scalable performance.

Qualifications

• Bachelor's or Master's degree in computing or related field (or equivalent).
• Proficient in Go, Python, or Shell scripting in Linux.
• Strong Kubernetes knowledge and container tech: Docker, Containerd, Kata Containers, Podman.
• Solid understanding of distributed systems, designing and maintaining large-scale distributed systems.

Responsibilities

• Design and develop resource scheduling systems for ML platforms, supporting training, evaluation, and inference workloads across NLP, CV, and Speech.
• Optimize orchestration and scheduling of GPUs, CPUs, and accelerators for multi-cloud resources.
• Develop scheduling solutions for compute, RDMA networking, and storage for large-scale distributed clusters.
• Design workload scheduling across data centers, regions, and cloud environments for online services and batch jobs.

Skills

Go

Python

Shell scripting

Kubernetes

Docker

Containerd

Podman

Distributed systems

Education

Bachelor's or Master's degree in Computing

Tools

Docker

Containerd

Kata Containers

Podman

Job description

ByteDance in Singapore is seeking a talented software engineer to design and develop resource scheduling systems for machine learning platforms, supporting training, evaluation, and inference workloads across NLP, CV, and Speech.

You will optimize the orchestration of GPUs, CPUs, and accelerators, and design workload scheduling across data centers and cloud environments to ensure efficient resource utilization and scalable performance.