AI Algo Eng - LLM/VLM Mandarin Required

Fuku; San Francisco (CA); Sur place; USD 170 000 - 250 000; Plein temps

Résumé du poste

Fuku in San Francisco is seeking an AI Algorithm Engineer to advance production AI systems, including LLMs, VLMs, and multimodal models. You will work at the intersection of research and engineering to shape content understanding, governance, and intelligent automation used by hundreds of millions of users.

The role emphasizes hands-on development, foundation model training, and agent architectures such as ReAct, PlanAct, and CodeAct.

Qualifications

• Master’s degree or above in CS/AI/ML/math or related.
• 3–5+ years of relevant industry experience.
• Professional Mandarin communication skills required due to collaboration with teams in China.
• Experience with LLMs, VLMs, agent systems, and multimodal models.

Responsabilités

• Develop production AI systems based on LLMs, VLMs, and multimodal foundation models.
• Design algorithms to understand signals across text, images, video, and audio.
• Build scalable multimodal intelligence for content understanding and decision-making.
• Improve semantic understanding, classification, retrieval, and reasoning across formats.
• Produce Agent systems including ReAct, PlanAct, CodeAct, and multi-agent architectures.
• Participate in foundation model training lifecycle (pre-training, SFT, RL).

Connaissances

Large Language Models

Vision Language Models

Multimodal Foundation Models

Agentic AI

Prompt Engineering

Retrieval-Augmented Generation

ReAct

PlanAct

CodeAct

Multi-Agent Systems

Context Engineering

Function Calling

MCP

A2A

RLHF / RL

SFT

Post-Training

Reward Models

Formation

Master’s degree or above in Computer Science, Artificial Intelligence, Machine Learning, Mathematics, or related disciplines

Description du poste

AI Algorithm Engineer – LLM/VLM (Mandarin Required)

Experience: 3–5+ Years
Education: Master’s Degree or Above
Employment Type: Full-Time
Sponsorship: H1B Transfers possible

About the Opportunity

Our client is a global leader in consumer internet platforms, serving hundreds of millions of users worldwide across content, community, and commerce. As the platform expands internationally, frontier AI is central to understanding, moderating, and improving large-scale multimodal content. We are seeking experienced AI Algorithm Engineers to develop the next generation of production AI systems, including Large Language Models (LLMs), Vision Language Models (VLMs), Agentic AI, and multimodal foundation models. This role involves working at the intersection of AI research and production engineering, directly influencing content understanding, recommendation quality, content governance, and intelligent automation at a global scale. This is not an LLM application or wrapper role.

Responsibilities

Foundation Models & Multimodal AI

• Develop production AI systems based on LLMs, VLMs, and multimodal foundation models.
• Design algorithms to understand complex signals across text, images, video, and audio.
• Build scalable multimodal intelligence for large-scale content understanding and decision-making.
• Enhance semantic understanding, classification, retrieval, and reasoning across diverse content formats.

Agentic AI

• Design and build production-grade Agent systems, including:
• ReAct
• PlanAct / CodeAct
• Tool Use & Function Calling
• Multi-Agent Architectures
• Context Engineering & Memory Management
• MCP / A2A
• Agent Orchestration
• Task Planning & Intent Understanding
• Productionise Agent systems to deliver measurable improvements to real-world products used by hundreds of millions of users.

Foundation Model Training

• Participate in the full model lifecycle, including:
• Pre-training
• Supervised Fine-Tuning (SFT)
• Reinforcement Learning (RL)
• Post-Training
• Improve model reasoning, planning, and multimodal capabilities.
• Design evaluation methodologies for models and Agent systems.
• Develop reward models and optimization pipelines for real-world production objectives.

Applied AI Research

• Evaluate emerging developments in LLMs, VLMs, Agentic AI, and multimodal learning.
• Translate frontier research into scalable production systems.
• Contribute to technical direction, architecture, and AI best practices.
• Opportunities to contribute to publications at leading AI conferences.

Requirements

• Master’s degree or above in Computer Science, Artificial Intelligence, Machine Learning, Mathematics, or related disciplines.
• Approximately 3–5+ years of relevant industry experience.
• Strong understanding of:
• Large Language Models
• Vision Language Models
• Multimodal Foundation Models
• Agentic AI
• Hands-on experience with:
• Prompt Engineering
• Retrieval-Augmented Generation (RAG)
• Agent evaluation frameworks
• Model evaluation
• Experience with modern Agent architectures, including:
• ReAct
• PlanAct
• CodeAct
• Multi-Agent Systems
• Context Engineering
• Function Calling
• MCP
• A2A
• Familiarity with:
• SFT
• RLHF / RL
• Post-Training
• Reward Models
• CodeRL or equivalent reinforcement learning infrastructure
• Strong engineering skills with the ability to translate research ideas into production AI systems.
• Professional Mandarin communication skills are required due to close collaboration with engineering and research teams based in China.

Preferred Experience

• Production experience with Vision Language Models or multimodal foundation models.
• Experience building production Agent systems at significant scale.
• Background in recommendation systems, search, content understanding, trust & safety, or content moderation.
• Experience with large-scale model training or post-training.
• Publications at conferences such as NeurIPS, ICLR, ICML, CVPR, ICCV, ACL, EMNLP, or KDD are advantageous.

Why Join

• Work at the intersection of frontier AI research and large-scale consumer products.
• Build production AI systems operating across one of the world’s largest content ecosystems.
• Engage with cutting-edge LLMs, VLMs, Agentic AI, and multimodal foundation models.
• See advances in model capability translate directly into impact for hundreds of millions of users worldwide.