字节跳动基于多模态大模型与用户行为理解的下一代分发与交互技术-抖音Generative Recommendation（北京/上海/杭州/深圳）

团队介绍：Data抖音推荐 Generative Recommendation 团队，通过对推荐模型的范式升级，带动整个推荐系统做整体重构，同时支持抖音短视频、直播等重要业务场景。我们的工作涉及大规模推荐算法的前沿探索、复杂约束的优化问题的解决、CV/NLP等多个学术领域的算法改进工作、面向工程约束下的最优算法设计和实现，以及对产品数据的复杂深入的分析工作。 1、结合业务目标和场景特点，利用并探索大型语言模型技术（LLM）对现有推荐系统的各个模块进行优化改进； 2、探索开发下一代生成式推荐系统技术，不断改善用户体验，实现业务目标。

We are looking for a Generative AI Intern Engineer to join the NVIDIA Developer Technology group (Devtech) and work with a team of experienced engineers on innovative uses of AI for games and content creation. The Devtech team works with NVIDIA researchers and leading game developers to bring cutting edge AI research from across NVIDIA and the industry to gamers and 3D professionals in high performance packages such as real-time inferenced graphics, physics and animations. What you’ll be doing: • Research and implement innovative generative AI algorithms for game engines and authoring tools, including real-time neural graphics, physics based animation and diffusion models. • Develop neural graphics, animation and physics models and maintain open-source projects for both game-making and user runtimes. Integrate them into mainstream game engines and DCC tools. • Use various optimization techniques, such as tensor fusion and quantization, to fit the AI models onto user devices and maximize the performance of inference for real-time gaming. • Collaborate with game developers on optimizations and improvements for specific GenAI applications. • Interact closely with the architecture and driver teams at NVIDIA in ensuring the best possible experience on current generation hardware, and on determining trends and features for next generation architectures.

Architect AI Agents – Design, orchestrate, and deploy robust AI agent architectures, including multi-agent systems, reasoning frameworks, and dynamic tool-use capabilities (RAG, APIs) to automate complex business workflows and analytics. Develop Agent Harnesses & Evaluation Pipelines – Build and maintain robust testing harnesses (Agent Harness), simulation environments, and CI/CD pipelines to rigorously evaluate agent performance, safety, reasoning logic, and tool-calling accuracy before production deployment. Collaborate & Innovate – Work closely with cross-functional teams to identify opportunities, gather requirements, and transform business challenges into scalable, agentic AI solutions. Build Robust Systems – Partner with data engineers and platform architects to implement high-performance, real-time GenAI applications. Ensure robust state management, secure memory handling, safety guardrails, and low-latency decisioning. Explore & Evolve – Stay at the forefront of LLM advancements and agentic frameworks (e.g., LangChain, LlamaIndex, AutoGen, Semantic Kernel). Continuously research and implement new paradigms in AI reasoning, planning, and evaluation to refine the team’s capabilities. Communicate Insights – Present complex agent architectures and GenAI strategies to business stakeholders and executives, ensuring technical innovations translate into meaningful business outcomes.

Engage with the latest BTP and Generative AI technologies, applying them to solve real-world business challenges Collaborate globally with cross-functional teams to design, prototype, and test innovative solutions within the BTP ecosystem. Participate in ideation sessions and contribute to planning of new technology implementations.