Algorithm Engineer – AIGC

Shopee; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

Shopee is seeking a highly skilled researcher to design and optimize training pipelines for large-scale AIGC models in a Singapore-based role. You will tackle diffusion-based multimodal generation tasks across image and video modalities, from architecture design to post-training alignment and quality enhancements.

The ideal candidate holds a Master’s degree and has top-tier publication credits, with strong proficiency in PyTorch, Python, C++, and CUDA, and excellent collaboration skills.

Qualifications

• Master's degree or above in Computer Science or related fields.
• First-author publication(s) at top-tier venues on diffusion-based multimodal generation.
• Experience with multimodal large models for image/video generation.
• Proficient in PyTorch.
• Strong programming in Python, C++, and CUDA; good collaboration/communication.

Responsibilities

• Model training & optimization for large-scale AIGC models, including diffusion models for visual tasks.
• Contribute to architecture and system design for multimodal image/video models.
• Advance post-training alignment methods to improve output quality.
• Explore capabilities like multi-modal understanding and visual content generation.
• Track frontier research, plan projects, and publish at top venues.

Skills

PyTorch

Python

C++

CUDA

Team collaboration

Education

Master's degree in Computer Science or related field

Job description

Job Description

• Model Training & Optimization: Design and execute training pipelines for large-scale AIGC models, including data curation, training stability, and performance optimization. Apply large-scale diffusion models to practical visual tasks such as low-light enhancement.
• Architecture & System Design: Contribute to the architecture and algorithm design for AIGC and large multi-modality models spanning image and video, from foundational model training through to post-training alignment and quality enhancement.
• Post-training & Alignment: Implement and advance post-training methods for diffusion-based multimodal large models to improve output quality.
• Capability Exploration: Investigate and prototype emerging capabilities such as multi-modal understanding and visual content generation.
• Frontier Research: Track cutting-edge AIGC research, drive project planning, and deliver production-grade implementations. Contribute to the academic community through publications at top-tier venues.

Requirements

• Master's degree or above in Computer Science or related fields
• First-author publication(s) at top-tier venues (CVPR/ICLR/ECCV/ICML/NeurIPS/TPAMI) on diffusion-based multimodal generation, particularly post-training/alignment methods that improve generation quality;
• Have experience in multimodal large model across image generation or video generation
• Proficient in PyTorch;
• Good programming skills in Python, C++, and CUDA;
• Strong problem-solving skills, innovative thinking, and excellent team collaboration/communication skills.