Agent感知算法工程师/专家 - PICO
Core
Build a first-person, long-term, continuous multimodal memory system for smart hardware using visual, audio, motion, and device state data to enable perception, understanding, and structured expression of user experiences.
Role type
Senior IC multimodal perception and memory algorithm engineer
Builds
Multimodal memory systems for XR smart hardware (glasses, wearables)
Domain
Consumer electronics / XR / AI
Deliverable
production ML models
Required skills
deep learning, multimodal modeling, first-person vision, long-sequence modeling, video understanding, behavior recognition, intent recognition, vector retrieval, cross-modal retrieval, information extraction, knowledge graphs, user profiling, model training, fine-tuning, evaluation, optimization
Preferred skills
smart glasses/wearable/robot product experience, long video understanding, first-person datasets, large-scale vector retrieval systems, on-device model deployment, inference acceleration, power and storage optimization, iterative optimization with online data
Technologies
deep learning frameworks, vector databases, knowledge graph tools, on-device inference engines
Responsibilities
Design memory systems for organizing, compressing, and mining massive multimodal data; extract high-dimensional user preferences and habits for search and recommendation; optimize multimodal understanding models to identify redundant/conflicting memories and extract key events; design algorithm solutions, prototype development, engineering integration, and performance optimization for product delivery
Seniority
Senior, hands-on IC
