AI Development Engineer

Job Responsibilities

1. Implement intelligent Agent systems based on large language models (LLMs) tailored to specific business scenarios.
2. Design and implement core modules of the Agent, including multi-turn dialogue state management, automatic task decomposition and planning (Task Planning), tool use strategies, multi-Agent collaboration mechanisms (role division/conflict resolution), and result evaluation and optimization, to ensure the system is "capable of thinking and acting".
3. Be responsible for the selection, fine-tuning, and inference optimization of LLMs (e.g., GPT-4, Claude, open-source models); build an efficient interaction layer between the Agent and external systems (API/database/third-party tool calls) to close the "perception-decision-action" loop.
4. Track cutting-edge technologies in the Agent field (e.g., ReAct, CoT, RAG, reinforcement learning applications), drive technological innovation based on business needs, continuously optimize system performance (response speed, task success rate, resource efficiency), and develop reusable technical components.

Job Requirements

1. Possess experience in the development of AI large model applications, and be familiar with the entire process of large model fine-tuning, RAG, inference and deployment.
2. Proficient in Python (a must); familiar with backend frameworks such as FastAPI/Flask, with capabilities in system design and code optimization; proficiency in Java/front-end technologies is preferred.
3. Skilled in frameworks such as LangChain/LlamaIndex, with experience in independently developing tool-use Agents or multi-Agent collaboration systems.
4. Candidates with experience in tool use or task planning (e.g., step-by-step reasoning based on LLMs) are preferred.
5. Master the core methods of Prompt Engineering (few-shot/CoT/ReAct prompting strategies) and be able to design efficient prompting strategies for different tasks.
6. Candidates with practical experience in the local deployment, quantization compression (INT4/INT8), inference acceleration (vLLM/TGI), or domain fine-tuning (LoRA/QLoRA) of open-source large models (e.g., Llama, Qwen) are preferred.
7. Have a strong passion for artificial intelligence technology, clear logical thinking, the ability to independently decompose complex problems, a strong desire to explore cutting-edge developments in AGI (e.g., AutoGPT, multi-Agent collaboration, Skills), and take the initiative to promote the implementation of technologies.