Technical Research
R&D History
Bairong began with Discriminant AI in 2014. By the end of 2017, it commenced research and development of Generative AI (specifically intelligent speech for multi-turn text and voice conversations). In 2018, this technology was successfully applied to intelligent operational services, resulting in the creation of the first-generation AI Chatbot product. Starting in 2021, it achieved large-scale commercialization.
Overall R&D Output
Since its establishment, Bairong Inc. has obtained over 260 enterprise software copyrights and patents, covering areas such as artificial intelligence, machine learning, privacy computing, human-computer collaboration, and multi-modal technologies. The company continues to attract and train top R&D talent, with nearly 50% of its employees engaged in R&D.
R&D Products/Tools
Bairong Inc. has developed efficient foundational tools tailored to specific business scenarios, including the ORCA automatic machine learning platform, BaiRong's large model BR-LLM, the intelligent AI voice product Chatbot, and the privacy computing platform Indra. These tools enhance product performance and facilitate efficient business upgrades.
BR-LLM Large Model
BaiRong Inc. has built the BR-LLM industrial large model based on the deep learning Transformer framework, combining NLP, Intelligent Voice Technology and other technologies. BaiRong Inc. independently developed the underlying framework for large models, enabling training with parameters in the billions through deep fine-tuning.
At the AI development level
large models demonstrate powerful code automation capabilities. Large model engineers can issue clear instructions in text to generate models, significantly improving machine learning development efficiency and production processes.
In intelligent interactions
large models offer "human-level" conversation effects, achieving voice recognition accuracy of over 99% for customers.
In decision analysis
small models, led by Discriminant AI, have already played a crucial role, and the introduction of large models will further stimulate the power of small models.
ORCA-GPT
utilizes BR-LLM's code generation capabilities to strengthen the foundation of ORCA, innovatively achieving automatic model conversion and deployment across programming languages and frameworks. This greatly reduces the development and deployment cycle and costs of model products.
C-EVAL
(a comprehensive Chinese basic model evaluation suite) ranks 25th in the latest test rankings. The model covers approximately 100 million high-quality Chinese training data tokens and has passed comprehensive tests from C-EVAL, including 52 different subjects and four difficulty levels. It demonstrates excellent performance in STEM, social sciences, and humanities, providing strong technical support for further exploring the specific applications of Generative AI technology in vertical fields.
AI Chatbot
integrates technologies such as natural language processing (NLP), automatic speech recognition (ASR), text-to-speech (TTS), and voice activity detection (VAD). It can accurately recognize human language, provide millisecond-level voice interaction and feedback, and achieve high accuracy. Currently, it is primarily used in the credit sector to assist institutions in tasks like activating existing credit card users, managing customer relationships in consumer loans, and exploring user value further.
