AI Researcher - Content Video Generation (Pre-training)

About the Company

Sea Group is creating a new strategic AI division focused on the next wave of generative AI. The mission is to expand what AI can do for human connection, self-expression, communication across languages, and richer social interaction. The team is also developing AI-native applications and a Model-as-a-Service (MaaS) support platform, built on large-scale multi-country data to help shape a multilingual AI ecosystem across Southeast Asia.

Within this broader effort, the AIGC team is working on advanced visual synthesis. Its goal is to lead the field in high-quality portrait and video generation through foundational research and the scaling of generative models for future social and e-commerce products.

Role Overview

This position is centered on large-scale pre-training infrastructure and distributed systems for AIGC models, with a strong emphasis on video generation workflows.

Responsibilities

• Develop distributed training toolchains that can support extremely large AIGC model training workloads.
• Improve distributed training efficiency across compute, network communication, and storage layers.
• Investigate and eliminate technical bottlenecks in training, with a focus on stronger stability and better throughput.
• Study emerging distributed training methods and take ownership of project planning as well as production-ready implementation.

Requirements

• A master’s degree in Computer Science or a closely related discipline is preferred; candidates with a bachelor’s degree may also be considered if they bring strong industry experience.
• At least 2 years of relevant professional experience is required.
• Solid understanding of distributed training concepts, including data, pipeline, tensor, and expert parallelism, along with practical experience applying them.
• Advanced working knowledge of major deep learning frameworks such as PyTorch, DeepSpeed, and Megatron-LM.
• Hands-on familiarity with GPU architecture and CUDA programming, including kernel development and debugging; knowledge of NCCL and cuDNN is also expected.
• Background in AIGC pre-training, Transformer-based models, and diffusion models such as Stable Diffusion and Flux.
• Strong analytical ability, creative problem-solving, and clear communication and collaboration skills.

Additional Information

This role is based in Singapore and is a full-time onsite position. The opportunity is part of a newly established AI department with a long-term focus on multilingual AI, generative systems, and production-scale model infrastructure.

Work Setting

Onsite in Singapore.