Build core Agent logic, including but not limited to task planning and orchestration, tool calling, multi-turn dialogue management, memory, RAG, context engineering, and multi-agent collaboration.
Lead Continuous Pre-training and Post-training for vertical domains and business scenarios, including building high-quality datasets and data pipelines, designing RL reward models, improving instruction following and reasoning capabilities, task completion, role-playing, anthropomorphic and personalized dialogue, proactive/reactive immersive multimodal conversation experiences, and enhancing the model's IQ and EQ.
Build long-term and short-term memory architectures, addressing issues such as forgetting and attention dispersion in long contexts, and improving immersion and consistency in long-term user interactions.
Build multimodal RAG systems, including development and optimization of key modules such as recall, ranking, long-text processing, and multi-document synthesis.
Develop the Agent's tool layer, integrating external APIs and MCP such as search, code interpreters, browsers, sandboxes, and third-party services.
Design and tune prompts and context management, with tailored optimization for different product requirements.
Design scientifically rigorous quantitative evaluation systems and plans aligned with product requirements continuously monitor product metrics and provide guidance for Agent and model optimization.
Explore innovative AI applications.
Requirements:
Master's degree or above in Artificial Intelligence, Computer Science, Mathematics, or a related field.
Strong programming skills proficient in PyTorch familiar with distributed training frameworks such as DeepSpeed and Megatron.
Strong experience with techniques and principles of LLM training/inference, including but not limited to data synthesis and filtering, model training optimization, prompt engineering, evaluation, deployment, and prototype development.
Strong problem analysis and resolution skills sustained interest and curiosity in frontier AI technologies and applications strong self-drive able to collaborate closely with teams to drive a full closed loop from research to deployment.
Good development experience with Agent frameworks such as LangGraph, Google Agent Development Kit, or OWL.
Strong engineering capabilities familiar with AI development tools such as Cursor and Claude Code proactive mindset for improving efficiency.
Prior project experience in areas such as Agentic reinforcement learning, virtual character generation, multimodal interaction, personality modeling, or memory will be a strong plus
