Case study · Climate tech
A live digital twinand AI control loop for carbon capture.
A U.S. company develops carbon capture and mineralization technology. Neuclon built the data and AI layer that connects its test unit to a live model of the process, and works with it as its long-term AI and data partner as it scales from pilot to commercial operation.
Active engagement · not yet a completed commercial deployment
- Industry
- Climate tech · carbon capture and mineralization
- Neuclon’s role
- Strategic AI and data partner
- Unit
- Three-section absorption column, solvent loop and reboiler
- Status
- Active engagement
The partnership
The AI and data foundation for a climate-tech platform.
Chosen as the AI and data foundation for the company’s climate-tech platform.
Showed how Neuclon supports its Design of Experiments process, turning experimental data into faster, more confident decisions.
Working alongside its commercialization strategy to help prepare the technology for scale.
Built to take in real-time sensor data and operate within hard, physics-defined safety bounds: the foundation for future autonomous control.
The challenge
Many interacting variables, one trade-off.
The test unit has a three-section absorption column, a solvent circulation loop and a reboiler that regenerates the solvent, monitored by 16 sensors. Raising the capture rate usually costs more energy, and the readings sat in separate instruments with no shared view of how the process was behaving.
What Neuclon built
From sensor to decision.
- Connected all 16 sensors through an edge gateway to a cloud IoT hub over encrypted MQTT.
- Streamed and validated readings in real time, and kept the full history in a time-series store.
- Built a live model of the unit, with every sensor mapped to its place in the process.
- Added physics-informed neural networks that respect the process chemistry and flow when predicting performance.
- Added a decision engine that recommends setpoint changes to raise the capture rate while cutting energy use, with confidence thresholds and engineer approval before anything changes.
How a reading becomes an action
- Unit sensors16 points over MQTT
- IoT hubDevice registry and telemetry
- Streaming ingestEvery reading as it happens
- Real-time validationBad readings caught
- Time-series historyThe full record
- Physics-informed AIPerformance prediction
- Decision engineConfidence and safety checks
- Approved setpointSent to the unit after approval
What’s next
From carbon to critical minerals.
The same physics-informed approach has been designed for critical-mineral extraction, searching process conditions far faster than running every experiment by hand.
The pattern carries straight into manufacturing: connect the sensors, build a trusted live model, and let AI agents recommend or make approved changes. It is how Neuclon works on a production line.
Talk about your process.
Bring one process where the trade-offs are hard to see. We will show you what a live model and an approved control loop would look like on your data.