Atomic-scale virtual reconstruction
Computable and verifiable models for molecular and materials systems under incomplete information, with explicit physical constraints and experimental alignment.
ML potentialsMD / REMDUncertainty
INDEPENDENT RESEARCH • COMPUTATIONAL SCIENCE • AI4SCIENCE
Reconstructing the microscopic world, one physical constraint at a time.
We develop computable, verifiable virtual reconstructions of atomic-scale systems — connecting machine learning, molecular simulation, scientific software, and experimental evidence.
Representation → Interaction → Exploration → Evidence
OUR PREMISE
When microscopic structure is incomplete, science does not have to stop. We build representations that can still be computed, evolved, challenged, and validated against physics and experiment.
RESEARCH
Three research directions, one reconstruction framework.
Our work moves from fundamental representation questions to industrially relevant chemistry and public-facing scientific tools.
Energy & chemical systems
Reaction pathways, degradation, interfaces, materials stability, and multicomponent systems — from high-accuracy electronic structure to scalable atomistic workflows.
MaterialsReactionsEnergy
Molecular science interfaces
Databases, web tools, educational software, and visual scientific experiences that translate rigorous microscopic models into usable public interfaces.
DatabasesWeb toolsEducation
APPROACH
From structure-driven simulation to information-driven reconstruction.
01
Representation
Encode molecules, compositions, structures, and partial observations into computable state descriptions.
02
Interaction
Model non-additive and cooperative effects using physics, data, and machine-learned potentials.
03
Evolution
Explore dynamics, reactions, thermodynamics, and alternative microscopic histories.
04
Validation
Confront predictions with uncertainty estimates, reference calculations, and experimental evidence.
HOW WE WORK
Research, collaboration, and tools.
22MI Research is designed as a small, high-agency research practice: deep technical work first, lightweight structure, and outputs that can become publications, prototypes, software, or collaborative programs.
R
Research programs
Independent and collaborative research in computational molecular science, advanced materials, and AI4Science.
Discuss a project →C
Scientific collaboration
Custom modelling, technical assessment, computational workflows, and research support for complex scientific questions.
Discuss a project →T
Tools & public science
Scientific databases, research software, interactive molecular tools, and rigorous learning experiences.
Discuss a project →ABOUT 22MI
An extra month for questions that do not fit neatly into existing boxes.
“Twenty-Second Month” does not exist on the calendar. The name represents a research space outside conventional boundaries — a place to explore scientific problems before they have settled into a standard discipline or workflow.
22MI Research is built around computational physics, chemistry, machine learning, and scientific software, with an emphasis on technically rigorous work that can connect to experiment and real-world use.
Physics firstModels should remain accountable to physical reality.
Evidence over spectacleClear validation matters more than impressive demos.
Build the interfaceResearch becomes more useful when it can be accessed, tested, and reused.
COLLABORATE
Bring us a difficult technical question.
We are interested in research collaborations, technical evaluation, computational modelling, and early conversations around emerging scientific technologies.
hello@22mi.exampleReplace with your production email before launch.