大模型记忆算法工程师（Agent/多模态方向） - PICO
Core
Develop algorithms for agent memory systems that extract information from dialogue, voice, images, and video to build event, entity relationship, task status, and long-term memory structures.
Role type
Senior IC machine-learning engineer (agent memory & multimodal)
Builds
Agent memory systems for XR platforms serving consumer and enterprise users
Domain
AI/ML, Large Language Models, Multimodal Understanding, XR
Deliverable
production ML models
Required skills
Machine learning, Deep learning, Natural language processing, Information retrieval, Multimodal understanding, Large language models, RAG, Agent systems, Vector retrieval, Sparse retrieval, Re-ranking, Python, PyTorch
Preferred skills
Long-term memory, Long context, Personalized agents, Temporal knowledge graphs, Continuous learning, Speech and vision fusion, Speaker recognition, First-person video understanding, Wearable devices, Retrieval model training, Memory strategy learning, Long-cycle evaluation datasets
Technologies
Python, PyTorch, RAG, Vector databases, Knowledge graphs
Responsibilities
Research hierarchical organization, compression, merging, updating, and forgetting mechanisms for memory; Develop task-oriented memory retrieval and usage strategies including query understanding and temporal reasoning; Establish data and evaluation systems for memory accuracy and consistency; Collaborate with product and engineering teams to deploy algorithms in real-world scenarios.
Seniority
Senior, hands-on IC
