Multimodal LLM Algorithm Engineer Graduate (Global E-Commerce, Knowledge Graph) - 2027 Start (PhD)

岗位职责 / Responsibilities

About the Team

We are part of the Global E-commerce Algorithm team, building multimodal content understanding, product understanding, semantic matching, pricing intelligence, and intelligent Agent capabilities for global content-commerce scenarios. Leveraging our massive-scale video content and global product data, we explore NLP, CV, Multimodal LLMs, Video LLMs, Multimodal Embedding & SID, Multimodal Reasoning, and Agent technologies to power product catalog construction, identical product matching, content-product linking, price comparison, AIGC content generation, and next-generation generative search and recommendation.

We are looking for talented individuals to join our team. As a graduate, you will get opportunities to pursue bold ideas, tackle complex challenges, and unlock limitless growth.

Successful candidates must be able to commit to an onboarding date by the end of the year. Please state your availability and graduation date clearly in your resume.

Responsibilities

1. Build global product catalog and content understanding systems, modeling products, videos, images, merchants, and brands with structured and semantic representations.

2. Develop algorithms for identical product matching, product/merchant/brand deduplication, and cross-lingual product aggregation, addressing multimodal matching, real-time aggregation, multi-granularity clustering, and semantic alignment.

3. Build pricing algorithms for price comparison, product pricing, price monitoring, and anomaly detection, supporting subsidy programs, price competitiveness analysis, and supply optimization.

4. Construct multimodal semantic links across video-product, product-product, and video-video relationships to support trend understanding, product mining, AIGC generation, and content-commerce supply optimization.

5. Explore next-generation generative search and recommendation through full-format representation learning and SID modeling across videos, livestreams, products, and queries.

6. Design and develop generative algorithms tailored for high engaging, high conversion e-commerce visual creatives, solving challenges in identity-preserving and aesthetic style consistency to produce production-grade intelligent background synthesis，virtual try-on, and localised lifestyle scene generation.

7. Explore business-facing Agents such as price comparison Agents, product analysis Agents, content AIGC Agents, and merchant operation Agents, leveraging RAG, tool calling, planning, reasoning, and automated evaluation.

8. Own the end-to-end algorithm workflow, including data construction, model training, evaluation, deployment, badcase analysis, and continuous iteration.

任职要求 / Requirements

Minimum Qualifications

- Individuals who are completing or have recently completed a PhD degree in Software Development, Computer Science, Computer Engineering, or a related technical discipline.

- Strong research or practical experience in NLP, CV, generative models (e.g. Diffusion，GANs，ControlNet) , multimodal learning, large models, or Agents;

- Proficiency in PyTorch / TensorFlow and hands-on experience in model training, tuning, and deployment;

Preferred Qualifications

- Experience with LLM/VLM, Embedding/SID, RAG, Tool Calling, or Agent Planning is preferred.

- Experience with distributed training, mixed precision, inference acceleration, TensorRT, or LLM serving optimization is a plus.

- Experience in multimodal retrieval, product understanding, content understanding, search/recommendation, pricing algorithms, intelligent Agents, or AIGC projects is preferred.

- Passion for frontier AI technologies; familiarity with Video LLMs, Multimodal Reasoning, vision-language alignment, long-video understanding, or generative search/recommendation is a plus.

- Publications, open-source contributions, or strong algorithm competition records are preferred. Strong communication, ownership, and a drive to create real business impact are highly valued.