（27届 ASP）S&R&A-风控大模型算法工程师

大模型算法研发 ：参与大模型（含多模态大模型）在风控场景下的应用研究，包括但不限于后训练（Post-training）方法，提升模型在虚假交易识别、异常行为检测、违规内容识别、广告作弊识别等任务上的表现。 内容风控 ：结合图文、视频等多模态信息，设计并优化违规内容（如虚假宣传、违禁品、低质内容等）识别模型。 广告风控 ：识别广告场景中的刷量、作弊点击、虚假转化等异常行为，设计有效的广告反作弊策略和模型。 行为风控与建模 ：设计并开发风控策略和检测模型，识别电商场景中的虚假交易、刷单等异常行为，保障用户和商家权益。 数据分析与特征挖掘 ：深入分析交易数据、用户行为数据、广告数据、商品商家信息及多模态内容数据，结合大模型与传统异常检测方法进行建模分析。 业务集成与监控 ：理解风控业务目标，参与建设异常监控体系，快速发现并定位风险。 前沿技术探索 ：跟踪大模型、多模态、强化学习、异常检测等领域的最新研究进展，推动算法创新在风控场景中的落地。 LLM Algorithm R&D : Participate in research on applying LLMs (including multimodal LLMs) to anti-fraud scenarios, including post-training methods, to improve model performance on fraud detection, anomaly identification, content violation detection, and ad fraud detection tasks Content Risk Control : Design and optimize models for detecting non-compliant content (e.g., false advertising, prohibited items, low-quality content) using multimodal signals such as text, images, and video Advertisement Risk Control : Identify abnormal behaviors in advertising scenarios such as click fraud, fake impressions, and false conversions, and design effective ad anti-fraud strategies and models Behavioral Risk Control & Modeling : Design and develop anti-fraud strategies and detection models to identify fraudulent transactions and abnormal behaviors in e-commerce scenarios Data Analysis & Feature Engineering : Conduct in-depth analysis of transaction, user behavior, advertising, seller/item, and multimodal content data, combining LLM-based and traditional anomaly detection methods Business Integration & Monitoring : Understand risk control objectives, help build robust anomaly monitoring systems, and quickly detect and address risks Frontier Technology Exploration : Stay current with the latest research in LLMs, multimodal models, reinforcement learning, and anomaly detection, and drive innovative algorithms into production

2027届及以后计算机、人工智能、数学等相关专业硕士/博士优先，本科优秀者亦可考虑。 具备较强的大模型研究能力，有 LLM/多模态大模型 预训练、后训练或 Agent 相关项目经验者优先。 熟悉强化学习基本原理，有 RL 相关论文发表或项目实践经验者优先。 机器学习理论基础扎实，熟练使用 PyTorch 等深度学习框架，具备独立复现论文和实验设计能力。 优秀的编码能力，有较强的工程实现能力。 有风控、内容安全、广告反作弊、推荐等相关领域实习或竞赛经验者优先。 注重细节，学习能力强，善于独立思考和快速试错。 良好的沟通和团队合作能力。 具备较强的自我驱动力与抗压能力，乐于接受挑战性工作。 Master's/PhD student (Class of 2027 or later) in Computer Science, AI, Mathematics, or related fields preferred; outstanding Bachelor's candidates also considered Strong research capability in LLMs/multimodal LLMs, with hands-on experience in pre-training, post-training, or agent-related projects strongly preferred Familiar with reinforcement learning fundamentals; publications or practical project experience in RL is a plus Solid foundation in machine learning theory, proficient in PyTorch, able to independently reproduce papers and design experiments Excellent coding and engineering implementation skills Prior internship or competition experience in anti-fraud, content safety, ad anti-fraud, or recommendation is a plus Detail-oriented, fast learner, able to think independently and iterate quickly Good communication and teamwork skills Strong self-motivation and resilience under pressure, willing to take on challenging work 加分项 有大模型后训练背景经验加分。 有风控实习经验加分