Forty Years
of Process
Data. New
Intelligence.
Most chemicals firms approaching AI start from a cold dataset. Paragon starts from the opposite position. Thirty-eight years of validated industrial operation across multiple licensee sites has generated the operating data — feed compositions, reaction conditions, yield profiles, NDOM trends, energy curves, maintenance cycles — that machine intelligence is most valuable trained on.
Four Deployed Applications
Process Optimisation in Real Time
Predictive models trained on validated sulphonation operating data anticipate yield deviation, NDOM excursion and energy-intensity drift before they occur — enabling closed-loop adjustments to feed rates, reaction temperatures and neutralization kinetics inside the operating envelope.
Yield deviation prediction
Closed-loop feed rate / temp / neutralization adjustment
NDOM maintained at 0.5% vs 1.5–2.0% market norm
Predictive Maintenance & Asset Reliability
Equipment vibration, thermal and process-stream signatures fed into anomaly-detection models forecast component failure ahead of breakdown. Combined with the novel sulphur platform's structural 36-month maintenance cycle, this targets zero unplanned downtime over a deployment life.
Vibration · thermal · process-stream signatures
Anomaly detection trained on validated industrial cycles
Zero unplanned downtime · structural 36-month maintenance cycle
R&D Acceleration & Pipeline Programme Design
Generative chemistry models and reaction-pathway simulation accelerate the lab-to-pilot phase of pipeline programmes — biogas → surfactants, algae cultivation optimisation, FHPC concentrate formulation. Computational pre-screening reduces the experimental burden by an order of magnitude versus conventional bench-led R&D.
Generative chemistry · reaction-pathway simulation
Biogas · Algae · FHPC Concentrates
Order-of-magnitude reduction in experimental burden vs bench-led R&D
Plant Engineering Productivity
AI-assisted plant engineering — generative design of equipment lists, automated P&ID quality-checking, AI-assisted procurement specification — accelerates LSTK delivery cycles. Combined with the 42-day commissioning record at Pithampur, this delivers consistently faster plant handover than the market norm.
Equipment list generation · P&ID QC · procurement spec
Accelerated LSTK delivery cycles end-to-end
42-day commissioning record · Pithampur
Three Thesis Principles
AI augments first-principles engineering. It does not replace it.
38 years of process chemistry knowledge is the foundation. AI accelerates and improves the application of that knowledge. It cannot substitute for it.
Data quality beats data volume.
38 years of validated industrial operation across multiple licensee sites. Feed compositions, reaction conditions, yield profiles, NDOM trends, energy curves, maintenance cycles — documented, not reconstructed.
AI deployment is a licensable capability.
The AI moat extends the IP moat. Licensees access not just the process technology but the intelligent operating layer built on top of it — extending the economic case for the Paragon licence.
Precision over marketing.
- We do not claim AI will replace process engineers.
- We do not claim machine learning models can design novel chemistries from first principles without human chemistry expertise.
- We do not claim AI-driven autonomy in safety-critical operations. We engineer for human-in-the-loop control.