Looking for Taiwanese partners in industrial AI.

Predictive maintenance for packaging machinery

We develop anomaly-detection and failure-prediction models for multihead weighers and packaging lines. The bottleneck is data: we need real operating history, not only models.

• PLC event and alarm logs with timestamps and fault codes
• Load-cell signals, weighing deviations, hopper and actuator behavior
• Vibration, temperature, current, motor and drive parameters
• Cycle times, recipe changes and operating conditions
• Maintenance tickets, service notes and component replacement history
• Known failures with approximate time, root cause and corrective action

1. 1Partner shares anonymized telemetry and service records.
2. 2Akira cleans and structures the data, builds the models.
3. 3Together we validate on historical or live data and define a measurable POC.

Few labelled failures? Still a fit. A first POC can start with anomaly detection on normal operation and add failure labels as service events accumulate.

MCP integration layer for AI agents

Not another chatbot. A ready control layer that lets AI agents use existing company systems through approved tools, so your team does not rebuild the integration stack for every customer.

• Connects Claude, ChatGPT and Cursor to ERP, MES, CRM, mail, files and databases through one gateway
• Selected functions exposed as tools, allowed or blocked per team and per user
• Every tool call recorded in an audit log: who, when, which tool, what result
• Users' system credentials encrypted with AES-256-GCM, never shown to admins or to the AI
• Optional pseudonymization of personal data before it reaches an external model
• Works with cloud models as well as private or on-premise ones