GitHub Scholar LinkedIn
001 / Archean · Berkeley Lab

Solving hard
problems with AI,
physics & chemistry.

I'm Théo Jaffrelot Inizan — founder at Archean, AI chemist & associate researcher at Berkeley Lab. I build AI that understands physics and chemistry, and point it at the hardest problems in matter.

Explore research Publications →
Founder, Archean · Associate Researcher, Berkeley Lab
drag to rotate
Materials generated by AI, synthesized in the lab.
T. Jaffrelot Inizan et al. — Nature, 2026
002 / About
Portrait of Théo Jaffrelot Inizan

I’m the founder of Archean and an Artificial Intelligence Chemist & Associate Researcher at Lawrence Berkeley National Lab. I build AI that understands physics and chemistry — and point it at the hardest problems in matter.

Previously, I was a researcher at UC Berkeley and Berkeley Lab, working with Omar Yaghi, Kristin Persson, Jennifer Chayes and Christian Borgs on generative AI — from diffusion models to LLMs and foundation atomistic models — to discover, design and understand materials science, chemistry and biology.

Before: PhD at Sorbonne University with Jean-Philip Piquemal, Fulbright scholar at UT Austin, M.Sc. at EPFL.

Bakar FellowSorbonne PhD ’23FulbrightEPFLUT Austin
UC BerkeleyLawrence Berkeley National LaboratorySorbonne UniversityEPFLUniversity of Texas at AustinNVIDIAAmazon Web Services

Research

003 / what I work on
— A

Generative AI for reticular chemistry

LLMs and diffusion models that propose, reason about and refine metal–organic frameworks — MOFGen: from prompt to synthesized crystal.

— B

Universal ML potentials

Machine-learned interatomic potentials for high-throughput prediction — phonons, stability, dynamics.

— C

Polarizable MD & HPC

Tinker-HP, adaptive sampling and hybrid neural/physics simulation at supercomputer scale.

Selected work

004 / highlights
01

MOFGen: generative AI designs synthesizable MOFs

Diffusion-designed frameworks, experimentally realized — AI-MOF-1–7.
Nat Commun · subm.
02

Large language models for reticular chemistry

Where LLMs take materials design next — with the Yaghi & Chayes groups.
Nat. Rev. Mater 03

Scalable hybrid deep neural networks/polarizable potentials biomolecular simulations including long-range effects

Neural potentials meet long-range physics at biomolecular scale.
Chem. Sci ’23 04

High-resolution mining of the SARS-CoV-2 main protease conformational space: supercomputer-driven unsupervised adaptive sampling

Unsupervised adaptive sampling of a pandemic target's conformational space.
Chem. Sci ’21

Recent publications

005 / 16 total
Large language models for reticular chemistry Nat. Rev. Mater · in press Machine Learned Potential for High-Throughput Phonon Calculations of Metal-Organic Frameworks arXiv · submitted AMOEBA Polarizable Molecular Dynamics Simulations of Guanine Quadruplexes: from the c-Kit Proto-oncogene to HIV-1 bioRxiv · preprint All publications & talks →

In the lab & beyond

006 / moments
Portrait Research event at UC Berkeley
007 / Contact

Don’t hesitate to reach out.

Always happy to talk AI, chemistry and materials — questions about MOFGen, the papers, or anything in between.

Google ScholarGitHubLinkedInX / Twitter