[Job Description]

This position is mainly responsible for tracking, reproducing and verifying the cutting-edge technology of AI Agent. You need to explore how to make large language models have stronger logical reasoning, tool calling and long-term memory abilities, and provide algorithmic feasibility support for product landing.

[Core Responsibilities]

1. Front-end technology research: in-depth analysis of the latest papers in the LLM Agent related field (such as Multi-Agent Collaboration, Planning, Memory Mechanism), and write a technical research report.

2. Algorithm validation: Design and validate Agent's Prompt Engineering strategy (such as CoT, ReAct) and workflow scheduling logic for specific business scenarios, and improve task success rate.

3. Data Construction: Responsible for building high-quality supervised fine-tuning (SFT) data or reinforcement learning (RLHF/DPO) preference data to optimize the model's performance under specific instructions.

4. Evaluation system: Build an Agent capability evaluation set, quantify model performance from dimensions such as accuracy, response speed, and security, analyze abnormal cases and propose improvement solutions.

[Job Requirements]

1. Educational background: Computer Science, Artificial Intelligence, Mathematics or related disciplines, Master's degree or above (outstanding undergraduate students may also be considered).

2. Technical skills: Proficient in Python, familiar with PyTorch/TensorFlow frameworks; familiar with the Hugging Face ecosystem.

3. Domain knowledge: In-depth understanding of Transformer structure, familiar with mainstream open-source models (such as Llama, Qwen, Mistral) and closed-source model API features.

4. Priority conditions:

   • Experience in using or reading the source code of open-source frameworks such as LangChain / AutoGPT / MetaGPT.