Skip to main content
A

Acaysia

Acaysia provides a drop‑in edge AI appliance that learns and continuously optimizes the control of chemical reactors in real time using physics‑informed machine learning and Model Predictive Path Integral optimization. The system integrates with existing PLCs via OPC UA or EtherNet/IP, delivers sub‑millisecond inference, and includes a millisecond‑scale failsafe that reverts to the plant’s PID controller within 100 ms to maintain safety.

Cambridge, United StatesFounded 2025350+ followers
Updated 2 months ago

Funding

Funding not disclosed

Funding rounds are not available yet.

Founders

Founder details are not available yet.

Product

Problem

Chemical manufacturers often rely on static PID loops and manual tuning for batch and continuous reactors, leading to suboptimal yields, excessive energy consumption, and delayed response to safety excursions. Integrating advanced control algorithms typically requires costly hardware replacements and extensive downtime.

Solution

Acaysia offers a drop‑in edge AI appliance that learns the dynamics of existing reactors and continuously optimizes control actions in real time. Using physics‑informed machine learning and Model Predictive Path Integral (MPPI) optimization, the system computes optimal trajectories under process constraints while the plant’s PID maintains steady‑state operation. A millisecond‑scale failsafe architecture automatically reverts to the proven PID controller within 100 ms on any fault, preserving safety instrumented system integrity. The appliance connects to standard PLCs (Rockwell, Siemens, Beckhoff, Schneider) via OPC UA or EtherNet/IP, requiring no production downtime. Operators access a transparent dashboard that shows setpoints, constraints, confidence levels, and provides audit‑ready logs for every decision.

Target Audience

Primary customers are process engineers and operations teams at chemical, petrochemical, and specialty manufacturing plants that operate batch, CSTR, or plug‑flow reactors and seek higher yield, lower energy use, and enhanced safety without replacing existing control hardware.

Features

  • Physics‑informed ML model that continuously learns reactor behavior from on‑premises time‑series data
  • MPPI control engine running on NVIDIA Jetson Orin edge compute, delivering sub‑millisecond inference and deterministic execution
  • Millisecond failsafe design with automatic reversion to existing PID control, meeting ASIL‑D safety standards
  • Drop‑in deployment via standard OPC UA and EtherNet/IP interfaces, compatible with major PLC vendors
  • Shadow, Advisory, and Closed‑Loop operation modes for staged validation and operator confidence building
  • Operator dashboard with real‑time setpoint visualization, constraint monitoring, confidence scores, and one‑click rollback
  • On‑premises data storage and secure export for model retraining and versioned rollback
This profile is AI-generated and may contain inaccuracies.