AI and computer vision platform for smart city infrastructure
中关村科学城城市大脑builds machine vision and AI systems for urban planning and smart city deployment. The tech stack is heavily weighted toward deep learning (TensorFlow, PyTorch, PaddlePaddle, Keras, Caffe) with emerging LLM integration (vLLM, Transformers, LangChain, RAG), suggesting a shift from pure vision models toward multimodal AI. The senior-heavy headcount (22 of 36) and focus on government-scale projects indicate a pre-sales and implementation-driven business model.
中关村科学城城市大脑operates a smart city platform combining machine vision, embedded algorithms, and cloud control infrastructure for municipal and urban planning customers in China. The product roadmap spans computer vision platform implementation, smart city solution design, and cloud control platform productization. The company employs 36 people across engineering (16), sales (7), product (6), data (3), operations (3), and research (1), with majority senior and mid-level IC roles. Pain points center on model optimization (lightweight models, hardware acceleration, small-sample learning), large-scale government project execution, and algorithm performance tuning.
Primary stack includes TensorFlow, PyTorch, and PaddlePaddle for model training, with emerging LLM tooling: vLLM, Transformers, LangChain, and RAG. Also uses Keras, Caffe, and MXNet for specialized vision and embedded workloads.
Active projects include smart city planning and solution design, machine vision platform implementation, cloud control platform productization, embedded algorithm optimization, and large-scale government project delivery.