美图Summer Intern- International Growth & Marketing (JP/KR/TH/VN/PT)
任职要求
If you’re a digital native living on the front lines of every trend, have a hound-like nose for hot topics, and can turn dry data into vivid user stories—stop scrolling! We’re not looking for "translators"; we’re looking for explorers who dare to "try" new ways to win global hearts. This is a full-time summer 2026 internship (5+ months). Strong performers will be considered for a return offer. Who We’re Looking For: Cultural Insider: You don't just speak Japanese, Korean, Thai, Vietnamese, or Portuguese; you live the culture. You know what makes local Gen-Z tick. Aesthetic Enthusiast: A power user of photo/video apps with a refined taste for beauty, fashion, or AI. 5G Digital …
工作职责
无
1、策划并执行电商平台级活动项目，包括大型购物节（如黑色星期五）、季节性促销（如夏季大促）及其他跨品类重点活动，推动全平台销售与生态增长； 2、管理端到端活动策略，涵盖促销补贴、卖家&创作者激励、内容增长及资源整合，统筹预算消耗与投资回报效率； 3、跨部门协同推进活动落地与优化，对接客户经理、达人经理及用户研究、公关、设计、产品、研发等职能团队； 4、基于活动数据表现分析，持续输出优化建议（如创新玩法、合作资源及提升活动效果的激活策略）。 1.Conceptualize, develop and execute e-commerce platform-level campaign projects, including major shopping festivals (e.g. Black Friday), seasonal campaigns (e.g. Summer Sale) and other major cross-category campaigns, responsible for platform-wide sales and ecosystem growth; 2.Manage end to end campaign strategies including promotion&subsidy, seller&creator registration, content growth and resource integration, manage budget consumption and ROI efficiency; 3.Collaborate and coordinate across stakeholders, including working with account managers, creator managers, and all other cross functional teams (e.g. user research, PR, designers, product team, R&D team etc) for campaign implementation and optimization; 4.Conduct data analysis on campaign performance and offer ongoing recommendations for campaign optimization, such as innovative ideas, partnerships and activations that enhance campaign performance.
-参与Qianfan-VL多模态模型训练工作，负责图像、视频与文本模态的对齐和指令微调数据构造与清洗，模型实验 -构建优质代码数据语料库，通过检索生成结合执行环境自动验证的方式构造合成数据，通过持续预训练、强化学习等方式训练精专代码大模型 -跟进并探索前沿的模型范式，研究样本质量/数量对模型效果的影响，摸索更高效、低成本的蒸馏方案，研究RL在特定需求场景下的适用性 -深刻理解模型训练、推理的性能瓶颈，通过底层优化提升训练、推理性能 -研究方向（新增数据策略方向）： 多模态大模型预训练/后训练数据清洗与合成 多模态推理场景强化 多模态教育、解题场景增强 复杂指令遵循 Agent场景数据构建 复杂场景下的代码提升
我们是百度智能云AI计算加速团队，负责大模型训练和推理框架的优化工作，为语言模型、多模态模型、强化学习等先进的AI技术提供超大规模、极致性能、兼容多芯的软件方案，支持百度内外部客户模型的快速迭代和高效部署。在这里，你可以： -紧跟最新LM技术的发展，从模型-框架-硬件集群的全局视角，优化分布式训练和推理框架、底层算子的性能，将常见模型训推MFU优化到极致状态 -在百度自研芯片上适配常见的大模型与自驾模型，结合芯片特点做框架和算子的深度优化，充分发挥自研芯片算力 -结合文生视频、具身智能、自动驾驶等前沿场景，模型-框架-集群协同创新，探索前沿的训推框架和算子优化技术 -与业务团队协同，理解客户需求，驱动技术创新，业务持续突破
-大模型训练、推理通信算子性能优化 -大模型算力集群网络性能优化 -大模型算力集群高可用系统优化 -大模型基础设施相关的其他优化