Since its release at the end of November 2022, ChatGPT, a large language model chatbot application launched by OpenAI, has rapidly aroused widespread attention worldwide.As a company specializing in artificial intelligence (AI), SenseTime has accumulated years of expertise in large models.SenseCore, a new generation of AI infrastructure computing cloud services, provides a three-in-one solution consisting of cloud-native AI infrastructure, an AI development platform, and AI expert services.With SenseCore as its core, our big model will completely change the development paradigm of AI and even the software industry, resulting in a profound impact to the business model, product design, and user experience of numerous industries in the future.SenseCore supports industries through its rich large model capacity, and empowers ecological cooperation to build a large model ecology that permeates the industry, both upstream and downstream.
Large-scale elastic cloud-native multi-level heterogeneous scheduling, product-level thousand-GPU support, maximum single-task training of 3200 GPUs; excellent resource utilization rate, lossless container support, high linear acceleration ratio of GPUs (>90%); ultra-high-speed and ultra-high-performance modern training network (total bandwidth approaching 100 Tbps, single node up to 800 Gbps); deeply optimized storage and data system: ultra-high IOPS (> 5 million) AI cache acceleration system, 1 million IPOS file storage, ultra-large-scale object storage system, small file-friendly, support for 500PB-level storage pool; support for PyTorch, TensorFlow, SenseParrots and other pairs of frameworks.
Full-stack tools for data, training and inference to deliver large model fine-tuning, automated model deployment and one-click inference services to accelerate large model R&D; AI-for-AI assists AI R&D with AI large models, initiating the AI R&D 2.0 paradigm.
Based on the profound professional experience in model development and training accumulated by SenseTime in the process of AI algorithm implementation, along with 20,000+ models trained across 20+ industries, it provides customers with large model algorithm development services, large model training optimization services and large model inference optimization services to help customers achieve the high-efficiency large model development and training and the low-cost deployment of applications.
⁃ Automatic annotation: 422% faster annotation
⁃ Hard Case mining: quickly solve the long-tail problem of autonomous driving
⁃ Small model production: low storage, high accuracy
⁃ Autonomous driving simulation: improve iteration speed and reduce cost
Large protein models are widely used in AI pharmaceuticals, AI biosynthesis and other fields. SenseTime cooperated with top biotech manufacturers in the industry to carry out antibody 3D structure feature extraction using the big model and antibody affinity prediction based on structure features, which led to a 60% improvement in experiment efficiency.
⁃ Digital human: professional explanation of financial knowledge, intelligent recommendation of financial products, and patient reply to business processes
⁃ SenseAuto Cabin: visual + voice + text emotion recognition, accurate judgment of the driver emotions
⁃ Intelligent recruitment: quick screening of talent, intelligent sorting of resumes combined with interview feedback
Provide basic computing power services of various specifications, including GPU and CPU computing power services, integrated storage services, high-performance network services, etc.
Include multi-faceted security services such as host machines, network layers, container layers, application layers, and data asset layers.
Provide an industrial-level AI R&D platform integrating professional AI tool chains and open AIalgorithms to realize end-to-end AI R&D services from training to inference.
Help you achieve new breakthroughs in business with professional AI solutions and advanced AI products
