AI Agent Algorithm Engineer

Job Overview

Join Shopee in Singapore as an AI Agent Algorithm Engineer to spearhead advancements in AI agent core logic, data pipelines, and multimodal conversational systems. This role focuses on developing innovative solutions in task orchestration, memory management, retrieval-augmented generation (RAG), and multi-agent collaboration, ensuring enhanced AI intelligence and engagement.

Core Responsibilities

• Develop foundational agent functionalities including multi-turn dialogue handling, task planning, tool invocations, memory modules, and inter-agent cooperation.
• Lead ongoing pre-training and post-training initiatives tailored to specific domains and business needs by crafting quality datasets, refining reinforcement learning reward models, and improving AI’s instruction adherence, reasoning, and interactive capabilities.
• Design both long-term and short-term memory architectures to counteract forgetting and attention loss in extended contexts, enhancing conversational immersion and continuity.
• Build and optimize multimodal RAG systems involving components such as data recall, ranking algorithms, processing lengthy documents, and synthesizing multi-document information.
• Integrate external APIs and multi-capability platforms including search functionality, code interpreters, browsers, sandboxes, and third-party services within the agent’s toolset.
• Create and fine-tune prompts and manage conversational context with tailored adaptations for varied product specifications.
• Establish rigorous, data-driven evaluation frameworks aligned with product objectives, regularly analyze metric trends, and provide actionable feedback for continuous agent and model enhancements.
• Investigate and experiment with cutting-edge AI application opportunities.

Candidate Requirements

• Master’s degree or higher in Artificial Intelligence, Computer Science, Mathematics, or a similar discipline.
• Minimum two years of professional experience designing and deploying multi-agent large language model systems focusing on task delegation, orchestration, and tool integration.
• Proven skills in fine-tuning large language models using supervised fine-tuning (SFT) and direct preference optimization (DPO), alongside expertise in building and refining retrieval systems involving recall strategies, ranking, and embedding tuning.
• Strong programming abilities, particularly proficient in Python.
• Exceptional analytical and problem-solving skills; demonstrated passion for emerging AI technologies; self-motivated and capable of collaborative full-cycle development from research to production deployment.
• Practical experience using agent development frameworks such as LangGraph, Google Agent Development Kit, OWL, or AutoGen.