Turing Technology was founded in 2014. We started with a self-developed database kernel and built a full stack from data foundation to models and knowledge to AI-native applications, with verifiable deployments in finance, industry, government, mobility, and energy.
2014 · Founding
The company was founded in 2014 and chose a full-stack, self-developed path from day one. That choice still shapes every product: we are not a systems integrator. Database kernels, storage engines, and other core components stay in our hands, so later AI capability sits on a foundation we control.
Building the stack
In this period we moved from investment to products and formed two core assets.
The first is Antelope Database. One kernel carries relational, vector, time-series, document, and graph data, with compute-storage separation, in-database AI, high reliability, and Xinchuang fit. It is in production at 70+ enterprises, including GF Securities, China Southern Airlines, and China Southern Power Grid. Kernels and storage engines are self-developed; the IP is ours.
The second is credentials and IP. We are a National High-Tech Enterprise; a Guangdong Specialized, Refined, Distinctive, and Innovative SME; and a Guangzhou Science and Technology Innovation Little Giant. We hold CMMI3, ISO 9001, and information-security management certifications, and have passed MIIT Xinchuang evaluation. We hold 90+ software copyrights and 30+ invention patents, and contribute to industry standards.
In the same period we assembled a core team spanning industry and academia: Zhang Yue, Executive Director and General Manager; Feng Lei, technical advisor (formerly Greenplum); Wang Yaqing, technical advisor (formerly Oracle and JD.com); and Zhou Tao, head of R&D (formerly Qi An Xin). Their work in distributed databases and information security is the technical base for what followed.
2022 · Scale in government mobility
The Foshan internet rental e-bike supervision platform went live—a complete delivery of our stack in government transport. It covers Chancheng, Gaoming, Sanshui, Nanhai, and Shunde. After three consecutive service phases it still runs with 125,000+ registered vehicles, 215,000 average daily orders, 92.6% parking compliance, 95.4% helmet compliance, and 97%+ vehicle online rate.
That launch proved a self-developed data foundation can carry city-scale real-time supervision. The platform later added a supervision agent for automated analysis reports and travel-pattern insights—from data supervision to intelligent supervision.
2025 · Breakthrough in core finance
GF Securities’ OTC derivatives AI trading-instruction module (Xinchuang edition) went live—a landmark in our finance work. It is deployed on the internal network with Xinchuang adaptation, turning IM-style spoken orders into execution for OTC derivatives, with 96% instrument recognition accuracy and 7×24 service.
In the same period we advanced institutional AIGC at GF Securities across 4 directions, 8 scenarios, and 45 requirements. Branch databases hold 400 TB and support 2 million concurrent queries with response under 2.3 seconds.
The point of this chapter: AI entered one of finance’s most unforgiving correctness regimes. The product locked a design rule that now governs every Turing agent: the large model may explain, not decide. The LLM understands and plans; a deterministic layer validates and executes; the counter interface is the only execution path.
2026 · AI-native application system
On top of the data foundation and model layer, the product shape moved from “data platform” to an AI-native application matrix. TuringFin research is commercial and live as SaaS for securities, funds, and asset managers. TuringRAG enterprise Q&A is commercial for enterprises, finance, and government. Contract-review agents and macro-cycle AI continue to expand.
In August 2026 we published an industry-facing overview of the AI-native stack—data foundation, models and knowledge, and AI-native applications—and stated the integrated claim: multimodal data, large-model scheduling, and agent collaboration.
The through-line
Product shapes have changed over more than a decade. The technical line has not: data → knowledge → action.
Antelope Database prepares the data. Multi-model scheduling and an industry semantic layer turn data into usable knowledge. Industry AI-native applications turn knowledge into action, and results flow back into the data system—a loop of sense, analyze, decide, and act. That loop is the moat: full-stack self-development, industry data that improves models, governable and auditable agents, and a fully domestic, Xinchuang-compliant stack.