AeroMuse灵感悬浮
中文

AeroMuse · Applied AI for industry · Guangzhou, China

AI that works
where the work happens.

We build on-premise LLM agents, computer-vision monitoring systems and digital workflow platforms. Knowledge stays inside the client's network, agents call only registered tools, and every output is traceable.

Explore solutions Contact us
04product lines
03systems running in real environments
300+staff using our OA system daily
01research desktop application in development

01What we doProduct lines

Four product lines,
from knowledge to execution

We turn large language models, multimodal understanding, retrieval-augmented generation, computer vision and workflow engines into systems that can be deployed, audited and iterated in production.

  1. 01

    Industry Agents & LLM Applications

    Offline / edge-deployed LLMs, multimodal courseware parsing, course knowledge bases with RAG, content review and guardrails, low-code agent workshop.

    • Edge LLM
    • Multimodal
    • RAG
    • Content review
    • Agent workshop
    CaseAI Smart Classroom
  2. 02

    Computer Vision & Smart Monitoring

    Multi-camera ingestion, vehicle load-state recognition, cross-camera similarity matching, human review loop, map alerts and statistics dashboards.

    • Multi-camera
    • Load state
    • Cross-camera
    • Human review
    • Dashboards
    CaseDumping Vehicle Monitoring
  3. 03

    Enterprise Digitalization & Collaboration

    Flowable-based approval workflows, web and WeChat mini-program clients, WeChat service notifications and SMS activation, organization and permission management.

    • Flowable
    • WeChat mini-program
    • Notifications
    • Permissions
    CaseOA for a Technical College
  4. 04

    Research & Engineering Software Intelligence

    Fixed workflows from theoretical calculation to simulation to report, agent tool calls via MCP, MATLAB / PSCAD connectors, template-based deliverables.

    • MCP
    • MATLAB
    • PSCAD
    • Fixed workflow
    • Word templates
    In developmentTheory–Simulation Assistant

02Selected workCases

Systems running
in real environments

Client names are anonymized to protect client information. All systems below have been delivered or field-validated.

Case 01Smart educationEdge deploymentField-validated

AI Smart Classroom for a Party School

An on-premise classroom agent for a provincial power-grid training academy. Courseware is parsed into a per-course knowledge base before class; during class, questions flow into the lecturer's private "parking lot" and the lecturer decides when to answer; after class, a one-page summary and FAQ are generated automatically. Everything runs inside the classroom intranet on a local compute server.

Client
Provincial power-grid training academy
Deployment
Classroom intranet + local compute server, no external access
Clients
Lecturer · learner · management
Key tech
Multimodal parsing · RAG · content review
Read case
Case 02Urban governanceIn production

Illegal Dumping Vehicle Monitoring

Multi-camera recognition of empty vs. loaded trucks, cross-camera similarity matching that triggers human review, and automatic evidence retention with suspected routes drawn on the map.

Read case
Case 03CollaborationIn production

OA System for a Technical College

A web + WeChat mini-program system used daily by about 300 staff, covering leave, document circulation, meeting rooms, official vehicles, seals and maintenance requests.

Read case

03How we buildPrinciples

Our product principles

I On-premise

Knowledge and data stay inside the client's network or edge nodes, with no dependence on external services.

II Controlled

Agents call only registered tools; no improvised scripts, no skipped steps; critical actions run only after user confirmation.

III Traceable

Every answer cites its sources; every action can be audited. Production paths fail loudly instead of returning mock results.

IV Non-disruptive

Systems embed into existing processes and connect to existing information systems; people keep their own rhythm.

Have a scenario in mind?

From requirements to on-site validation and ongoing iteration, we work alongside our clients until the system is genuinely usable. Tell us about the business, the environment and the boundaries.

Contact us hejianhao1989@aeromuse.cn