Computational Separation Science

Redesigning how rare earths are separated

AI models that learn the physics of solvent extraction, designing separation cascades from specifications instead of trial and error.

Learn More Get in Touch
The Rare Earth Elements
17 elements. Nearly identical chemistry.
Scroll
The Problem

China controls 90% of rare earth processing. America has no alternative.

Every F-35, every EV motor, every wind turbine depends on rare earth elements processed almost exclusively in China. The US has the ore in the ground. What it lacks is the technology to turn that ore into usable materials without sending it overseas. That's a national security vulnerability, and it starts with separation.

0
US facilities producing
separated heavy REEs at scale
$0M+
estimated cost to build a
single separation facility
3–5 yrs
typical timeline to design
and commission a plant
0
chemically near-identical
elements to separate
0+
extraction stages in
a single cascade
Our Approach

Physics-informed deep learning

Instead of months of physical experimentation, we train deep learning models on the governing physics of solvent extraction, learning the mappings from equipment geometry to separation performance.

How Solvent Extraction Works
Chemically near-identical elements are gradually separated across hundreds of mixer-settler stages
Feed
Mixed REEs enter
Extract
Solvent grabs targets
Scrub
Impurities removed
Strip
Product recovered
Pure
≥ 99.5%
La
Ce
Nd
Pr
Dy
01
Forward Surrogate
Geometry-aware models that predict separation performance from equipment specifications in milliseconds.
02
Inverse Design
Given target purity and recovery, automatically design the machine geometry that achieves it.
03
Circuit Optimization
Bilevel optimization across entire cascade flowsheets, where the architect designs and the builder executes.
Vision

From specifications to separation, automatically

We're building toward a world where designing a separation process is as simple as specifying what you need. Tell the system your target purities, feedstock composition, and constraints. It designs the plant.

Vorelios Designer
Feedstock
Target Element
Target Purity
—
Optimize For
Min Cost
Min Footprint
Max Throughput
Cascade stages
Est. recovery
Design time
Design Cascade →
Contact

Let's talk critical minerals

Whether you're in defense, clean energy, mining, or government, we'd like to hear from you.

contact@vorelios.com