Al Engineer, Large Language Model

Shopee; Singapore; On-site; SGD 120,000 - 170,000; Full time

Job summary

Shopee's Compass team in Singapore is seeking a PhD-level researcher to contribute to pre-training, alignment, and evaluation for ultra-large multilingual models. You will work on Mixture-of-Experts, RLHF, and safety improvements, applying your deep knowledge of Python and PyTorch to experiments and model training.

Collaborate with a cross-disciplinary team to push progress from research to business-ready AI solutions, sharing results and exploring new directions for enhancement in real-world

Qualifications

• Doctorate required: Computer Science, Information Technology, Programming & Systems Analysis, or related discipline.
• Excellent coding skills in Python; strong data structures & algorithms knowledge.
• Minimum 1 year of research in training methods for industry-leading LLMs.
• Experience in text generation or dialogue systems.
• Strong problem analysis and collaboration skills.

Responsibilities

• Contribute to pre-training and alignment algorithm research for ultra-large multilingual models.
• Advance safety and explainability for trustworthy LLMs.
• Benchmark frontier technologies and assess applicability to business scenarios.
• Design and run experiments to test model performance and identify improvements.
• Collaborate with a team using statistics, scripting, and Python/PyTorch.

Skills

Python/PyTorch

Data structures & algorithms

Research experience in LLM

Text generation / dialogue systems

Communication & collaboration

Education

Doctorate degree in CS/IT/related

Job description

About The Team

Compass is Shopee's globally competitive, ultra-large-scale multilingual foundation model, built with advanced cross-cultural understanding, complex reasoning, and agentic execution capabilities. Leveraging the rich multilingual and multicultural data advantages of global emerging markets — particularly Southeast Asia — the Compass model family has successfully evolved from v1 through v3. We are now pushing the frontiers of foundation model research to build the next-generation base model with more advanced architectures and superior capabilities, establishing a world-leading technology system for multilingual intelligence and Agentic LLM. Through deep research into algorithms and data strategies, we aim to develop principled scaling laws to guide model planning, break through training and inference efficiency bottlenecks, and forge a deterministic path from data governance to emergent intelligence.

Job Description

• Contributions to the research and implementation of pre-training and alignment algorithms include ultra-large-scale multilingual pre-training technology, Mixture-of-Experts model training, Instruction Pretraining, SFT, and RLHF.
• Contribute to the explanation and safety improvement of AI, especially in trustworthy Large language models.
• Follow the frontier technologies and make comparisons about the advanced technologies to apply in business scenarios.
• Conduct experiments to test the performance of different AI models, identifying areas for improvement and exploring new directions for enhancement.
• Work collaboratively in a team environment, applying expertise in statistics, scripting, and relevant programming languages.

Requirements

• Doctorate degree in Computer Science, Information Technology, Programming & Systems Analysis, or other related disciplines
• Excellent coding skills, data structure and basic algorithm skills, proficiency in Python/Pytorch coding.
• Minimum 1 year of research experience in basic principles and training methods of industry-leading LLM (such as GPT, LLaMA).
• Have research experience in text generation or dialogue systems
• Excellent problem analysis and solving skills, able to deeply solve problems in large model training and application.
• Good communication and collaboration skills, able to explore new technologies with the team and promote technological progress.