JR0288261 AI Agent 实习生（AI 软件优化方向）

AI Agent 核心框架研发与性能优化 负责 AI Agent Runtime、Agent Kernel 及任务执行框架的设计与优化，研究智能体在复杂任务场景下的规划、推理、工具调用与执行机制，持续提升任务完成效率、稳定性与可扩展性。 Agent 系统基础能力建设 围绕 Agent Workflow、Memory、Tool Use、Multi-Agent Collaboration 等核心组件开展研发工作，优化上下文管理、任务编排、状态维护及执行链路，提升智能体系统在复杂环境中的自主执行能力。 Agent 性能分析与工程化落地 构建 Agent 系统的性能评测、Profiling 与监控体系，定位并解决复杂场景下的性能瓶颈，推动 Agent 技术在研发效率提升、自动化运维及智能工程等场景中的规模化应用

基本要求 计算机科学、人工智能、软件工程或相关专业的在读硕士或博士研究生 熟练使用 Python，具备良好的软件工程习惯和扎实的代码能力 对 深度学习系统或大模型推理 至少一个方向有深入理解，包括但不限于： PyTorch / TorchDynamo / FX / Inductor 大语言模型推理流程（如 KV Cache、Prefill / Decode） 对大语言模型（LLM）的理解不仅停留在 API 使用层面，能够理解其推理过程、性能瓶颈和工程约束 具备较强的问题拆解能力，能够将复杂工程问题抽象并建模为 AI Agent 可执行的任务结构 （二）加分项 具备大模型量化相关经验，如 GPTQ、AWQ、SmoothQuant、MXFP、FP8 等 熟悉主流模型推理加速或部署框架 了解或实践过高性能计算、底层算子或内核相关开发与优化 有 AI Agent 实际落地经验（不限学术背景），如 Agent 架构设计或工具调用实践 具备复杂工程系统的开发或维护经验，如自动化测试、评测、CI 或性能分析工具 有开源项目贡献经验（模型、系统、编译、加速或 Agent 框架相关均可） 可连续实习 4 个月及以上 ，每周线下到岗 4 天及以上 AI Agent Intern (AI Software Optimization) Job Responsibilities AI Agent Core Framework Development and Performance Optimization Design, develop, and optimize AI Agent runtimes, agent kernels, and task execution frameworks. Research planning, reasoning, tool invocation, and execution mechanisms for agents in complex task scenarios, continuously improving execution efficiency, system reliability, and scalability. Agent Core Infrastructure and Capability Development Develop and enhance core Agent components, including workflows, memory systems, tool usage, and multi-agent collaboration. Optimize context management, task orchestration, state management, and execution pipelines to improve autonomous operation and decision-making capabilities in complex environments. Agent Performance Analysis and Production Deployment Build performance evaluation, profiling, and monitoring frameworks for Agent systems. Identify and resolve performance bottlenecks in large-scale and complex workloads, and drive the adoption of Agent technologies in areas such as software engineering productivity, intelligent operations, and automated engineering workflows. Qualifications Basic Requirements Master’s or PhD student in Computer Science, Artificial Intelligence, Software Engineering, or a related field Proficient in Python with solid coding skills and good software engineering practices In-depth understanding of at least one area in deep learning systems or large model inference , including but not limited to: PyTorch / TorchDynamo / FX / Inductor LLM inference workflows (e.g., KV cache, prefill / decode stages) Understanding of large language models beyond API-level usage, including inference behavior, performance bottlenecks, and engineering constraints Strong problem decomposition skills, capable of modeling complex engineering problems into executable AI Agent task structures Preferred Qualifications Experience with large model quantization techniques Familiarity with mainstream inference acceleration or deployment frameworks Knowledge of or experience in low-level kernel, operator, or performance optimization Hands-on experience in deploying AI Agents for real-world tasks Experience with complex engineering systems such as automated testing, benchmarking, CI, or performance analysis tools Open-source contributions related to models, systems, compilers, acceleration, or Agent frameworks Availability for an internship of at least 4 months , with on-site work 4 days per week or more var box = document.getElementsByTagName('body')[0]; window.parent.iframeHeight(box.clientHeight)