News Center

BidiTech has launched AquaBrain, an on-premises large language model platform for water utilities, treatment plants and pump stations. It brings a private LLM, an expert knowledge base and operations analytics together in one console, with all data and models running on the customer’s own servers.

Why an on-premises water LLM

Operating data and internal documents are sensitive, and general-purpose cloud models neither understand water engineering nor meet compliance needs. AquaBrain keeps the model, vector search and knowledge base in-house, with separate workspaces for each group and plant.

Four core capabilities

• Water knowledge base: procedures, standards, equipment documents and textbooks, with formulas and tables parsed faithfully and every result traceable to its source.
• Knowledge assistant: answers grounded in the knowledge base with citations; when no source exists it says so.
• Operations diagnosis: structured output covering symptoms, likely causes, checks, actions, and safety/escalation.
• Operations analytics: an analytics engine computes energy, specific energy and pump-combination efficiency exactly; the model interprets the results and writes the report.

In practice: three improvement opportunities from one analysis

At a lift pump station, AquaBrain analysed about 26,000 minute-level records from five pumps over 18 days and produced a report within minutes. The most efficient pump combination used about 13% less energy per cubic metre than the everyday combination but ran less than 4% of the time; two fixed-speed pumps showed a power factor of only 0.71–0.76; and a wrong frequency scaling factor was detected in one pump’s data acquisition.

From knowledge to control

AquaBrain is designed alongside BidiTech’s MPC precision control, rooted in 40 years of research at Shanghai Jiao Tong University, and its AIPC edge controllers. Next, validated control models will be deployable to plant-side edge controllers, and inspection data from BidiTech’s own acoustic camera will flow directly into AI for spectrum-level fault classification.

Utilities and partners are welcome to book a demo using their own operating data. Learn more at biditech.cn/solutions/water-llm.

In large enterprises’ digital transformation, a “digital vacuum” has long existed between top-level ERP/MES and shop-floor physical control. Fragmented processes, dormant data, heavy manual work, and weak system alignment constrain efficiency and cost.

BidiTech bridges hard real-time control and business operations with a dual-engine architecture—industrial physical simulation plus AI Agents—acting as the “intelligent glue” between the physical and digital worlds.

1. AI Agents: a 24/7 intelligent hub

An AI Agent is not mere automation: it is a digital brain that operates autonomously across systems, closes end-to-end loops, and both decides and executes.

• Wake up enterprise data; boost retrieval efficiency by about 80%
• Cut onboarding and Q&A costs by about 60%
• Parse 100+ document formats; RAG with traceability to reduce LLM hallucination
• One agent spans R&D diagnostics, audit Q&A, marketing attribution, finance automation, and more

2. Five scenarios where agents deliver value

Tender intelligence agent: 24/7 opportunity capture and LLM-based matching—customers include leading water-equipment firms.

Omnichannel marketing ROI agent: unify Xiaohongshu, JD, Meituan, Dianping, etc.; auto ROI attribution and live dashboards—serving a major OEM.

Finance & order automation: capture → verify → bank reconciliation → smart dunning. Month-end: 5 days to 8 hours; ~70% labor reduction—finance and F&B chains.

Tax & invoicing compliance: ERP to tax systems; e-invoices and paper in parallel; verification and deduping—error rate down to ~0.1%, proactive compliance—leading agri-food group.

Import/export customs: email listening → parsing → document prep → clearance—~500% faster customs work, >98% document accuracy—cosmetics and logistics multinationals.

3. Dual engine: deep control, intelligent operations

Hard control layer: decades of world-model and MPC heritage (SJTU lineage); AIPC edge controllers and DeepControl for real-time performance under strong coupling and long dead time.

Agent layer: IPA + AI runs business flows across systems, linking control, execution, and management.

5. What sets us apart

• Full stack in-house: algorithms, hardware, agents—no patchwork
• Fast rollout: sidecar integration, 2–6 weeks, minimal disruption
• National credentials: high-tech enterprise, dozens of invention patents
• Nationwide delivery: Shanghai HQ, Hangzhou R&D, Shandong manufacturing, Suzhou services

BidiTech—industrial physical control and AI Agent solutions: connect physical and digital, close the intelligence gap, and make every device, process, and system work in sync.

On March 17, the CPC Xinzhuang Town Committee in Minhang hosted the session “Pooling XIN momentum · Smart manufacturing in Xinzhuang — Party-building for the industrial software ecosystem,” successfully concluded. Under the theme of Party leadership empowering industry, BidiTech joined the regional ecosystem with strong technical support for industrial digital transformation.

Inside the ecosystem: energizing the “red chain”

The town inaugurated a Party-building council for the industrial software ecosystem. BidiTech answered the call for shared resources and complementary strengths to help build a full regional industrial-software value chain.

Advisory role: BidiTech’s CEO received the “Industrial software ecosystem expert advisor” certificate. BidiTech will contribute to policy dialogue and digital transformation roadmaps for the region.

Collaboration: through platforms such as “XIN Smart Industry Hub,” BidiTech will work with partners on key technologies, homegrown industrial software, and AI synergy.

Technical depth with the innovation center

During the event, Technical Lead Dr. Ni Zhenjiang visited the Minhang Industrial Digital Innovation Center for in-depth exchanges.

AI Agents and world models: Dr. Ni presented foundational work on industrial world models—embedding physical constraints so robots and AI gain “physical intuition” for richer prediction and perception in complex conditions.

Operational “physical rapport”: BidiTech’s industrial AI Agent stacks enable deep software–equipment interoperability; both sides aligned on using this rapport to advance digital transformation in industry.

Scaling impact: the teams discussed deploying AI digital systems in water, intelligent manufacturing, and other process industries, turning pilots into replicable industrial benchmarks.

BidiTech: physical intelligence for manufacturing

BidiTech is an industrial AI solution provider turning cutting-edge AI into productive outcomes. Its core team combines decades of industrial experience. DeepControl controllers and AIPC edge devices integrate APC and large-model forecasting, delivering cost, efficiency, and control wins for listed leaders— with flagship deployments in water, smart manufacturing, pharma, and more across China.

Closing

From going alone to pooling momentum, BidiTech will keep empowering manufacturing through indigenous innovation under the ecosystem. Closer ties with the Minhang Industrial Digital Innovation Center will add practical know-how for domestic industrial software.

Website: www.biditech.cn