Backend Engineer for AI Platform & Large-Model Infra

BYTEDANCE PTE. LTD.; Singapore; Presencial; SGD 120.000 - 180.000; Jornada completa

Descripción de la vacante

ByteDance Pte. Ltd.

in Singapore is seeking an experienced engineer to design and develop core pipeline components for the Ark MaaS platform, supporting API capabilities for text conversations and multimodal generation. You will participate in evolving cloud-native architectures, including Service Mesh, load balancing, and routing, while ensuring high availability and low-latency performance for large-model multimodal tasks.

Formación

• Minimum qualifications: B. Sc or higher in Computer Science or related field with at least 3 years of relevant experience.
• Familiar with distributed systems on Linux.
• Proficient in multiple languages (Golang, Python, C/C++, Java, Scala, JavaScript). ACM ICPC/Codeforces winners preferred.

Responsabilidades

• Design and develop core pipeline components for Ark MaaS platform (Model as a Service) including API capabilities like text conversations and multimodal generation.
• Contribute to evolutions of cloud-native architectures (Service Mesh, LB, routing) and implement highly available solutions.
• Optimize system performance for large-model multimodal inference with high throughput and low latency.
• Ensure stability under large-scale model invocations and address architectural bottlenecks during traffic surges.

Conocimientos

Golang

Python

C/C++

Java

Scala

JavaScript

Distributed systems

ICPC/Codeforces

Kubernetes

Docker

Istio

Envoy

Educación

B. Sc or higher in Computer Science or related field

Herramientas

Linux

Descripción del empleo

ByteDance Pte. Ltd.

in Singapore is seeking an experienced engineer to design and develop core pipeline components for the Ark MaaS platform, supporting API capabilities for text conversations and multimodal generation. You will participate in evolving cloud-native architectures, including Service Mesh, load balancing, and routing, while ensuring high availability and low-latency performance for large-model multimodal tasks.