AI大模型算法专家LLM Algorithm Expert

任职要求

<p>职位描述</p><p>.主导大模型的算法研究、训练调优及工程化落地，提升模型性能和效率；</p><p>.主导大模型压缩、分布式训练、推理加速等技术落地（如量化、MoE、FlashAttention等）；</p><p>.结合业务场景（如对话系统、内容生成、知识推理）设计模型优化方案，解决数据稀疏性、幻觉抑制等挑战；</p><p>.跟踪学术界与工业界最新进展，推动技术成果转化；</p><p>.主导技术方案输出，协同工程团队实现高性能服务部署；</p><p>.参与Agent架构设计，完成整体Agent项目落地。</p><p><br></p><p>任职资格</p><p>.硕士及以上学历，2年以上大模型研发经验，5年以上算法研发经验；</p><p>.精通PyTorch/TensorFlow框架，熟悉分布式训练工具；</p><p>.深入掌握Transformer架构及衍生技术，有丰富的大模型调优经验；</p><p>.熟练掌握大模型微调、蒸馏、强化学习等训练技术；</p><p>.具备完整的大模型训练-部署全链路经验；</p><p>.熟练掌握Agent架构；</p><p>.有顶会论文发表者优先。</p><p><br></p><p><strong>Job Description</strong></p><ul><li><strong>LLM R&amp;D &amp; Deployment:</strong> Lead algorithm research, training, fine-tuning, and engineering implementation of Large Language Models (LLMs) to enhance model performance and efficiency.</li><li><strong>Optimization &amp; Acceleration:</strong> Drive the implementation of LLM compression, distributed training, and inference acceleration techniques (e.g., quantization, MoE, FlashAttention).</li><li><strong>Domain Adaptation &amp; Challenge Mitigation:</strong> Design model optimization strategies tailored to business scenarios (e.g., conversational AI, content generation, knowledge reasoning) to address challenges like data sparsity and hallucination mitigation.</li><li><strong>Tech Transfer:</strong> Track frontier developments across academia and industry to drive the commercialization of technical achievements.</li><li><strong>Architecture &amp; High-Performance Deployment:</strong> Lead the output of technical architecture solutions, collaborating with engineering teams to deploy high-performance model services.</li><li><strong>Agent Architecture:</strong> Participate in Agent system architecture design and drive end-to-end Agent project execution.</li></ul><p><strong>Qualifications</strong></p><ul><li><strong>Education &amp; Experience:</strong> Master's degree or above; 2+ years of LLM R&amp;D experience, with 5+ years of overall algorithm research and development experience.</li><li><strong>Frameworks &amp; Infrastructure:</strong> Proficient in PyTorch/TensorFlow frameworks and experienced with distributed training tools.</li><li><strong>Core Architecture &amp; Fine-Tuning:</strong> Deep mastery of Transformer architecture and related variant technologies, backed by extensive experience in LLM tuning.</li><li><strong>Training Techniques:</strong> Proficient in LLM fine-tuning, knowledge distillation, and reinforcement learning techniques.</li><li><strong>Full-Lifecycle Experience:</strong> Proven track record spanning the entire end-to-end pipeline from LLM training to deployment.</li><li><strong>Agent Expertise:</strong> Fluent in AI Agent architecture principles and design.</li><li><strong>Publications (Preferred):</strong> First-author or major publications in top-tier conferences are preferred.</li></ul>