GE医疗
AI Algorithm and Development Software Engineer
AI Algorithm and Development Software Engineer
发布于 2 天前普通员工/个人贡献者
北京市
高级经验
全职员工
仅现场办公
硕士
PyTorch
LLM
Prompt Engineering
RLHF
Fine-tuning
Chain Of Thought
AI 估算 · 30k–50k
AI算法高级岗位，北京一线城市，大型跨国企业，硕士博士学历要求，市场竞争力强，薪资处于较高水平。
职位详情
关于这个职位
该职位专注于大型语言模型（LLM）的核心技术研发与落地，包括模型微调、Prompt Engineering、Function Calling、Agent系统构建等
你将参与从算法研究到工程实现的全流程，推动AI技术在医疗领域的创新应用
适合有深厚NLP/深度学习背景、对LLM前沿技术充满热情的技术人才
最低要求
MS or PhD degree or above in Computer Science, Artificial Intelligence, Mathematics, Statistics and other related majors, with solid theoretical foundation in natural language processing, machine learning and deep learning.
More than 3 years of relevant working experience in large language model algorithm research and development, familiar with the training, fine-tuning and inference process of mainstream open-source large models (such as LLaMA, Qwen, ChatGLM series, etc.).
Proficient in deep learning frameworks such as PyTorch, TensorFlow, familiar with model fine-tuning tools such as PEFT, LoRA, and have hands-on experience in large model parameter-efficient fine-tuning.
In-depth understanding of Prompt Engineering, Function Calling, Chain of Thought and other LLM related technologies, with practical project experience in model reasoning optimization and long context processing.
Experience in building LLM evaluation systems and conducting model hallucination, stability and tool call accuracy testing is preferred.
Proficient in Python programming, with good data structure and algorithm foundation, and strong code implementation and problem-solving abilities.
Have strong learning ability and innovative thinking, pay attention to industry cutting-edge dynamics, and have the ability to independently tackle key technical problems.
Good communication and collaboration skills, able to efficiently cooperate with cross-team members to promote project progress.
工作职责
Be responsible for the fine-tuning of large language models, including supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), and domain-specific model adaptation, to improve model performance and adaptability in vertical scenarios.
Conduct in-depth research and optimization on Prompt Engineering, design high-efficiency and high-robustness prompt templates, and explore advanced prompt strategies to enhance model output quality and task completion efficiency.
Optimize the Function Calling capability of large models, improve the accuracy, stability and generalization of model tool invocation, and realize the seamless connection between models and external tools and services.
Conduct research and implementation on advanced model reasoning technologies, including Chain of Thought (CoT), reflection mechanism, multi-step reasoning, and long context management, to solve complex reasoning tasks and extend the effective context window of models.
Build a comprehensive LLM evaluation system, conduct all-round testing and evaluation on models, focusing on model hallucination problems, output stability, tool call accuracy, reasoning ability and other core indicators, and put forward targeted optimization plans.
Track the cutting-edge research progress and technical trends in the field of large language models, introduce advanced algorithms and technologies into business scenarios, and promote the continuous iteration and upgrading of model technology.
Be responsible for the research and development of the core Agent runtime system, including the design and implementation of execution engine, state machine, memory module and tool call framework, to ensure the efficient, stable and scalable operation of the Agent system.
Develop and optimize multi-Agent collaboration mechanisms, realize core functions such as dialogue routing, conflict resolution, task decomposition and aggregation, and build a collaborative system for efficient interaction and task division among multiple Agents.
Optimize the scheduling logic and execution efficiency of the Agent engine, solve the problems of task delay, memory overflow and tool call failure in the Agent operation process, and improve the overall performance of the system.
Design the Agent system architecture with high availability and high scalability, support the access of various types of large models and external tools, and meet the needs of complex business scenarios.
Be responsible for the connection and secondary development of mainstream open-source frameworks in the field of Agent, including LangChain, AutoGPT, OpenCWA, etc., and independently develop customized Agent frameworks and components according to business needs.
Collaborate with product, engineering and other teams to translate algorithm research results into implementable technical solutions, and support the landing and application of large model products.
Participate in agile processes: planning, estimation, retros, and on-call (as needed)
优先资格
CUDA/CuFFT/CuBLAS, OpenMP, SIMD vectorization
Experience with git, jenkens, devops, etc.
Publications/competitions or open-source contributions
AI 洞察
优缺点分析
优点
- 前沿技术栈：直接接触LLM、Agent等最热门AI领域，技能积累价值极高
- 公司平台优势：GE医疗为全球领先医疗科技公司，资源丰富，有完善的研发体系
- 行业前景：AI+医疗是高速增长赛道，职位稳定性与长期发展前景好
- 国际化环境：有机会与全球团队协作，提升跨文化沟通能力
- 技术难度高：涉及RLHF、多Agent协作等复杂技术，需要较强的研究能力和工程能力
- 竞争激烈：该领域人才需求旺盛，但相应的高水平候选人众多，竞争较激烈
缺点 / 挑战
- 工作强度可能较大：涉及on-call，且需要跟踪前沿动态，持续学习压力大
- 适合对LLM技术有强烈兴趣、具备扎实AI算法背景和工程实践经验的求职者，喜欢挑战性工作并希望在该领域深耕的人
角色解读
- 技术深耕：从算法工程师向AI专家/首席科学家发展，专注LLM前沿研究
- 架构方向：转向AI系统架构师，负责大型AI平台的设计与落地
- 管理方向：晋升为技术团队负责人（Tech Lead）或AI团队经理，带领团队攻关
- 负责大语言模型（LLM）的微调与优化，包括SFT、RLHF和领域自适应，提升垂直场景性能
- 深入研究Prompt Engineering，设计高效提示模板，探索高级提示策略以改善模型输出质量
- 构建并优化Agent运行时系统，包括执行引擎、状态机、记忆模块和工具调用框架，支持多Agent协作
- 搭建LLM评测体系，针对幻觉、稳定性、工具调用准确性等核心指标进行全面测评并提出优化方案
- 精通PyTorch、TensorFlow等深度学习框架，熟悉PEFT、LoRA等微调工具，有实际大模型微调经验
- 深入理解Prompt Engineering、Function Calling、思维链（CoT）等LLM相关技术
- 熟悉Agent框架（如LangChain、AutoGPT）并能进行二次开发和自定义框架设计
- 扎实的Python编程能力和数据结构算法基础，具备独立解决复杂技术问题的能力
申请策略
- 申请前深入了解GE HealthCare在AI医疗方面的具体产品与解决方案，面试时展示对业务场景的理解
- 提前准备一个端到端的LLM项目案例，从微调、评测到部署的全流程，体现工程落地能力
- 突出在大语言模型微调（SFT/RLHF）方面的项目经验和具体成果，如模型性能提升指标
- 展示Prompt Engineering、Function Calling或Agent系统设计的实际案例，说明你在其中的角色和贡献
- 强调技术深度：如深入理解Transformer架构、熟悉多种开源模型（LLaMA、Qwen等）的内部机制
- 如有开源贡献或相关论文发表，务必在简历中显著标注
- 补充Agent框架的实战经验，特别是LangChain和AutoGPT，可以自己搭建一个简单的Agent demo
- 强化LLM推理优化技术，如CoT、长上下文处理，可以复现相关论文或参加Kaggle比赛
面试指南
- STAR原则（情境-任务-行动-结果）来回答项目经验问题，突出自己的技术决策和量化成果
- 对于设计类问题，先明确需求背景，再提出整体架构，然后分模块阐述技术选型和理由，最后总结优缺点和优化方向
- 请详细描述你做过的一个LLM微调项目，包括数据准备、微调方法、效果评估及遇到的挑战
- 如何设计一个高效的Prompt？请举例说明Chain of Thought Prompting的实现和优势
- 解释Function Calling的工作原理，以及如何确保模型工具调用的准确性和稳定性
- 请介绍你对Agent架构的理解，如何设计一个高可用的多Agent协作系统？
- 如何评价一个LLM的模型幻觉？有哪些方法可以缓解幻觉问题？
- 复习LLM基础知识，包括Transformer结构、预训练、微调、RLHF等核心概念，并能清晰解释
匹配度报告
74
综合匹配度
顶尖AI技术岗，超大平台高成长，工作强度一般，生活平衡需权衡。
适合人群
最适合强烈追求技术成长、希望从事AI前沿研发、接受一定程度工作强度且看重社会价值的求职者。
最强匹配
成长发展匹配
最弱匹配
工作生活匹配
薪资福利70
成长发展90
工作生活50
使命价值85
薪资福利匹配
70中等
薪资水平预计较高，但JD未明确具体薪酬福利，只提及提供搬迁协助。作为上市公司，薪酬体系规范但无额外福利亮点。
薪资信号未披露（AI估算：30K-50K/月）
福利待遇Relocation Assistance
成长发展匹配
90较高
职位涉及LLM最前沿技术，技能成长空间极大。JD明确提到了多种先进技术和开源框架，并有论文/竞赛加分项，发展导向很强。
技术前沿前沿/新兴技术
技术栈LLM、SFT、RLHF、Prompt Engineering、Function Calling、Chain of Thought、Agent、LangChain、AutoGPT、PyTorch、LoRA、PEFT
业务类型profit_center
工作生活匹配
50较低
JD未明确工作模式或WLB，但提及on-call要求，可能面临一定加班。办公地点在北京，未明确具体位置，通勤不确定。
工作模式仅现场办公
办公地点未明确
加班情况未提及（无法判断）
使命价值匹配
85较高
GE医疗致力于‘无限制医疗’，社会价值高。AI算法在医疗领域的应用直接改善患者体验，使命感较强。同时行业处于高速增长期。
行业发展高速增长赛道
社会影响正向社会影响力较高
使命信号create a world where healthcare has no limits、Every difference builds a healthier world
创新程度积极采用新技术
Watch Jobs