Agent Application Algorithm Engineer

About The Team

Sea Group is launching a pioneering AI division concentrating on generative AI to transform human interaction, self-expression, diversity in communication, and social engagement. This team is committed to creating advanced AI-native applications and a robust Model-as-a-Service ecosystem utilizing extensive multinational data, aiming to establish a premier multilingual AI platform in Southeast Asia.

The AI application division integrates social connectivity with artificial intelligence, aiming to develop digital personas serving as personal assistants and societal connectors. Functioning with startup agility yet leveraging the group’s substantial backing, the team is positioned to redefine human-AI interaction.

Job Description

• Enhance the quality of AI-generated content by improving multimodal comprehension of visual, audio, and video data; refining image and video generation workflows; and resolving issues related to output consistency.
• Develop core Agent functionalities covering task planning, orchestration, API/tool integration, multi-turn dialogue handling, memory systems, Retrieval-Augmented Generation (RAG), context engineering, and collaboration among multiple agents.
• Lead ongoing pre-training and post-training efforts tailored to specific domains and business applications, encompassing dataset creation, reward model design, improved instruction and reasoning capabilities, task execution, anthropomorphic and customized dialogue, and immersive multimodal communication featuring both proactive and reactive elements to boost model intelligence and emotional quotient.
• Construct short-term and long-term memory frameworks to mitigate forgetting and attention challenges in extended interactions, enhancing user experience consistency and immersion over time.
• Design and optimize multimodal RAG solutions, focusing on recall, ranking, handling long texts, and synthesizing multiple documents efficiently.
• Develop an Agent tool layer integrating third-party APIs and middleware platforms for search, code execution, browsing, sandboxes, and other external services.
• Create and refine prompt designs and contextual management strategies with customization for varying product demands.
• Establish rigorous, quantitative evaluation frameworks aligned with product objectives; continuously track performance metrics and provide actionable feedback for Agent and model improvements.
• Investigate and innovate with new AI applications.

Candidate Requirements

• At least two years of practical experience in computer vision, multimodal AI, or generative AI, including delivering production-level projects. Familiar with key vision-language models such as Qwen-VL, InternVL, GPT, Gemini series, plus diffusion and video generation techniques.
• Master’s degree or higher in Artificial Intelligence, Computer Science, Mathematics, or related disciplines.
• Proficient programming skills with expertise in PyTorch and familiarity with distributed training frameworks like DeepSpeed and Megatron.
• Solid understanding of large language model training and inference processes, data synthesis, filtering, training optimization, prompt engineering, evaluation, deployment, and prototype creation.
• Excellent problem-solving ability, continuous enthusiasm for cutting-edge AI technology, strong self-motivation, and ability to collaborate across teams to execute research-to-deployment cycles fully.
• Experience with Agent development frameworks such as LangGraph, Google Agent Development Kit, or OWL.
• Strong software engineering skills with knowledge of AI development tools like Cursor and Claude Code, and a proactive approach to enhance efficiency.
• Additional experience in Agentic reinforcement learning, virtual character creation, multimodal communication, personality modeling, or memory systems is highly desirable.