理想汽车Multimodal AI Agent Algorithm Engineer
任职要求
· Master’s degree or above in Computer Science, Artificial Intelligence, Natural Language Processing, Machine Vision, Speech Signal Processing, Mathematics, Physics, or related fields. PhD graduates are preferred (exceptions can be made for candidates with outstanding abilities). · Technical proficiency in deep learning frameworks such as PyTorch/TensorFlow, as well …
工作职责
· Responsible for the exploration and implementation of intelligent agent algorithms for the smart cockpit at Li Auto, promoting their application in scenarios such as dialogue systems, decision-making control, and multi-task interaction, while continuously enhancing the user’s intelligent experience. · Explore the design of agent architectures driven by large models, including but not limited to task planning, memory management, tool invocation, multimodal perception, and reasoning. · Lead the research, development, and optimization of algorithms that combine large models with intelligent agents, optimizing the integration of large models with reinforcement learning, knowledge graphs, environmental simulators, etc., to enhance the autonomy and generalization ability of intelligent agents.
关于我们： 我们是一支全球化、多元化、专业化的数据先锋团队，以技术为引擎，以数据为纽带，驱动全球20亿消费者与数千万商家的数字化商业生态。立足中国，服务全球，每天处理覆盖东南亚、欧洲、美洲等多时区的跨境数据洪流，在多语言、多文化、多法规的复杂场景中，打造“数据&AI技术驱动业务”的全球化数据中台。 团队致力于构建全新的满足安全合规的国际化大数据架构体系；统一的用户/商品/商家资产体系建设，含统一的DMP和选品平台；面向海外商家数据服务的生意参谋及数据银行支撑业务全链路数据驱动闭环，打造从站外竞对机会发现到商品供给和用户增长的数据智能服务Agent平台。我们秉承简单开放、创新能力、匠心精神的团队文化； 职位描述 Job Description 1. 深入理解行业业务逻辑与用户生命周期，通过用户行为分析、消费心理建模、多源数据融合，诊断业务增长瓶颈，设计可落地的用户价值提升策略（如会员分层运营、场景化精准触达、流失用户挽回等）。 2. 主导端到端增长项目：独立完成从业务需求拆解->实验设计->用户特征工程->预测模型开发（如客户分群/LTV/传播裂变因子挖掘）->策略效果归因的全流程。 3. 搭建业务分析框架：结合行业特性（如电商高频转化、内容平台沉浸度驱动、金融行业信用风险维度），设计可解释的用户标签体系与归因模型，输出用户洞察报告指导产品迭代与运营策略。 4. 与搜索推荐、产品、运营团队紧密协作，推动增长实验、A/B测试落地，结合AI模型结果，持续优化产品与内容分发策略。 5. 支撑用户增长策略的算法能力沉淀与平台化建设，推动AI在个性化推荐、多模态建模、用户行为预测等方向的深度应用。 1. Drive business growth strategies through deep user analytics and lifecycle value modeling, focusing on solving real-world problems like member tier operation, scenario-based engagement, and churn recovery. 2. Own full-cycle projects from business diagnosis to deployment: 3. Develop industry-specific frameworks: Design interpretable user tagging systems and attribution models tailored to sector characteristics (e-commerce conversion loops, content engagement drivers, etc.) 4. Collaborate closely with Search & Recommendation, Product, and Operations teams to run growth experiments and optimize strategies based on AI insights. 5.Contribute to platform-level capability building for scalable, AI-powered growth solutions across personalization, multi-modal modeling, and user behavior prediction.
1、建设AI原生智能研发产品，包括但不限于AI Coding IDE、研发 Agent、 代码大模型等，利用技术手段提高研发人员的开发效率。 2、利用Agent、RAG、Multi-Modal 等 AI 技术，打造 AI 程序员，覆盖前端、后端、测试等各种研发工种。 3、打造智能化的需求分析、估价、分发、调度系统，让人类与 AI 程序员高效协同，低成本高效率的完成各类复杂任务。
1、负责语音理解和语音生成算法在滴滴场景的落地使用 2、跟进最新技术，结合业务场景，提升语音识别、音频事件检测、声纹识别、语音合成等算法效果 3、探索语音大模型或多模态大模型在语音理解及语音生成场景的应用范式 4、算法优化，从模型架构、推理框架、量化压缩等角度提升模型推理速度、降低推理成本 Job Description 1. Responsible for the implementation of speech understanding and speech generation algorithms in Didi’s business scenarios. 2. Stay updated with the latest technologies and improve the performance of algorithms such as speech recognition, audio event detection, speaker recognition in real-world applications. 3. Explore the application paradigms of large language models or multimodal models in speech understanding and generation scenarios. 4. Optimize algorithms by enhancing inference speed and reducing costs through improvements in frameworks and quantization
团队介绍：BandAI团队致力于探索智能的极限在交易场景的可能性。团队研究方向涵盖LLM、Multimodal、Agent，在北京、上海设有实验室和岗位。加入我们，参与到前沿的大语言模型的研究课题，和优秀的研究员一起探索智能极限。 1、成为研究型人才，在你热爱的课题方向上，探索多模态大模型模型领域最具挑战的长期关键问题； 2、探索研究多模态理解、生成式、强化学习、AIGC等前沿技术； 3、探索多模态RAG、视觉COT、多模态Agent、多模态Reward model、RL等多模态进阶能力； 4、探索多模态Deep research、Computer Using Agent、Useful Image Generation、理解生成一体模型在抖音交易场景的能力。