WHY TEAMS CHOOSE PHIN

Stop guessing and start engineering.

Materials development has outgrown its tools, and that requires a new approach. PHIN unifies all your R&D: project planning, physical experiments, digital experiments, and uncertainty-aware quantum-grounded machine learning into one platform, giving your team the tool to to address your hardest problems.

Accurate across materials

Accuracy verified across chemistries, phases, and conditions

15x acceleration

Target for catalyst R&D under DOE ARPA-E, with PNNL, WashU, and Lectrolyst

R² > 0.99

Uncertainty aware models eliminate hallucinations

10,000x

Lower-cost screening than traditional quantum workflows

INDEPENDENTLY VALIDATED

Selected By The U.S. Department of Energy.

ARPA-E selected PHIN into its CATALCHEM-E program alongside Pacific Northwest National Laboratory, Washington University, and Lectrolyst, to compress industrial catalyst development to a single year.

Read the announcement
ARPA-E CATALCHEM-E award announcement animation
PUBLIC CASE STUDIES

PHIN transforms scientific development.

Both programs go beyond incremental improvement to show the transformative nature of PHIN. Autonomous agents pair with digital experiments to screen thousands of candidates a month, integrating directly with robotic and laboratory experiments to reduce decades of R&D and $100 million budgets by orders of magnitude.

DOE ARPA-E · CATALCHEM-E

Industrial catalysts

PHIN's agentic simulation platform connects to PNNL's MIRAL autonomous laboratory, with catalyst manufacturing and testing at Washington University and Lectrolyst, to develop heterogeneous catalysts for syngas and alcohol production.

IN-HOUSE · CASE STUDY

MOF polymer membranes

PHIN is developing a metal-organic framework polymer membrane that lets desalination plants operate at half the pressure, significantly reducing their electricity demand.

WHAT IS DIFFERENT

Six reasons teams pick PHIN over the alternatives.

Approve spend before every experiment

Agents scope the experiment and price it, then wait for a scientist to sign off, rather than spending first and reporting afterwards.

Manage all R&D in one place

Planning, physical and digital experiments, and the models behind them sit in one system, so program knowledge accumulates in the platform, not in the scientist you might lose.

Upload historical data

The runs you already paid for, including the failures, tune models to your chemistry, so you start from your lab's evidence rather than a public benchmark.

Spend lab time only on vetted predictions

Uncertainty quantification flags where a model extrapolates, so you catch an untrustworthy number before it costs a synthesis run. Generic potentials ship no such estimate.

Make techno-economic decisions

Machine-learning surrogates, physics simulation, DFT, and lab experiments each carry a known cost and confidence, so you escalate only when the answer is worth the price.

Compound every data point

Your teams capture each result once, and it tunes the models behind the next question, so the program compounds instead of restarting with every project.

Run PHIN on the materials problem your team keeps deferring.