Agentic Commerce AI Agent Algorithm Engineer

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

Job summary

Sea Group is building a new AI department focused on multilingual models and MaaS. We seek an experienced ML/NLP engineer to design a consumer-facing Shopping Agent, develop post-training pipelines, and optimize dialogue accuracy and product relevance.

You will collaborate with product, engineering, and operations teams to deploy capabilities at scale, with emphasis on data alignment, evaluation, and cost efficiency.

Qualifications

• Master’s degree or higher in CS, AI, or related field.
• 3+ years in ML/NLP, search/recommendation, or LLM apps.
• Hands-on with post-training methods such as SFT, DPO, PPO/RLHF, or Reward Modeling.
• Familiar with Megatron-LM, veRL, or DeepSpeed and basic distributed training.
• Experience with Agent, Function Calling, Tool-use, multi-turn dialogue, and LLM evaluation.
• Strong Python, engineering, data-analysis, and production problem-solving skills.
• Experience in e-commerce AI or large-scale post-training is a plus.

Responsibilities

• Design and optimize a consumer-facing e-commerce Shopping Agent covering intent understanding and dialogue.
• Develop and implement post-training pipelines including SFT/DPO/PPO-RLHF and Reward Modeling.
• Improve tool-use accuracy, instruction following, product relevance, and factual consistency.
• Build post-training datasets and evaluation systems with human and automated metrics.
• Drive continuous optimization via offline evaluation, A/B tests, and latency/cost improvements.
• Collaborate with product, engineering, search/recommendation, and operations teams to deploy Agent capabilities.

Skills

Python

NLP

Distributed training

Post-training methods

LLM evaluation

Problem-solving

Education

Master’s degree in CS/AI

Tools

Megatron-LM

DeepSpeed

veRL

Job description

About Us

Sea Group is establishing a brand-new, strategic AI department. This department is dedicated to exploring the transformative potential of generative AI in revolutionizing human connection, self-expression and communication diversity, and social interaction. We are building the next generation of AI-native applications and a comprehensive Model-as-a-Service (MaaS) product support system. Based on massive multi-country data, we are building a leading multilingual AI ecosystem from the ground up. We look forward to more outstanding talents joining us to build leading Southeast Asian multilingual models and explore innovative AI-native applications.

The AI application team focuses on the intersection of social connectivity and artificial intelligence. Our mission is to leverage LLMs to create digital personas that can act as personal assistants and social bridges. This team operates with a startup's agility backed by our Group's robust resources, aiming to define how humans interact in the AI era.

Job Description

• Design and optimize a consumer-facing e-commerce Shopping Agent, covering intent understanding, task planning, query rewriting, product search/recommendation, tool calling, memory, and multi-turn dialogue.
• Develop and implement post-training pipelines such as SFT, DPO, PPO/RLHF, and Reward Modeling, covering data construction, training, evaluation, deployment, and regression.
• Improve tool-use accuracy, instruction following, product relevance, factual consistency, personalization, and hallucination control through data and model alignment.
• Build post-training datasets and evaluation systems combining human evaluation, automated metrics, and LLM-as-a-Judge.
• Drive continuous optimization based on offline evaluation, A/B testing, user experience, conversion, latency, cost, and stability.
• Collaborate with product, engineering, search/recommendation, and operations teams to deploy Agent capabilities in large-scale e-commerce scenarios.

Requirements

• Master’s degree or above in Computer Science, AI, or a related field
• At least three years of experience in machine learning, NLP, search/recommendation, or LLM applications.
• Hands-on experience with at least one post-training method, such as SFT, DPO, PPO/RLHF, or Reward Modeling.
• Familiarity with training frameworks such as Megatron-LM, veRL, or DeepSpeed, and basic knowledge of distributed training.
• Experience with Agent, Function Calling, Tool-use, multi-turn dialogue, and LLM evaluation.
• Strong Python, engineering, data-analysis, and production problem-solving skills.
• Experience with e-commerce AI, large-scale post-training, Agent RL, self-play, or synthetic data is a strong plus.