Request for Proposals: AI for Actionable Matter Modeling

This is a pilot program. We are exploring whether AI-driven methods can reliably predict the material and molecular properties that determine technological function in the real world—defects, interfaces, disorder, and dynamics operating at multiple length and time scales. If this pilot uncovers compelling evidence, it may unlock a significantly larger, multi-year investment in this space.

This is a lightweight application. There are no letters of support, no spreadsheets to be filled. The application consists of a short online form and a (less than) 5-page project narrative pdf. We're interested in what you've built and where you want to take it—not your ability to write long proposals.

Overview

Fields that manipulate matter—chemistry, materials science, biology—face a persistent gulf between computational models and the lab bench. Conventional simulation emphasizes idealization: perfect crystal structures, static snapshots, vacuum environments, and single length or time scales treated in isolation. But the functional properties of foundational technologies (catalysts, semiconductors, drugs, batteries) are determined by non-idealities—defects, interfaces, disorder, dynamics—operating across scales that conventional methods cannot connect. As Nobel Laureate Herbert Kroemer remarked of semiconductors, "the interface is the device."  

We believe that AI may now help close this gap—not only by modeling non-idealities more accurately at a single scale, but by bridging scales that have historically required separate communities, codes, and assumptions. We are equally interested in methods that improve fidelity within a single scale and methods that connect scales that today are treated in isolation. We want simulations to move from post-hoc explanations to credible guidance that experimentalists use to decide what to synthesize or measure next.

This RFP is for you if:

• You are developing AI methods to predict non-idealities (defects, interfaces, disorder, dynamics) and/or multi-scale phenomena in technologically relevant materials or molecules
• You have a specific materials or molecular system in mind where conventional simulation underperforms
• You are at a university, national laboratory, or nonprofit research institution anywhere in the world

This RFP is not for you if:

• Your work primarily accelerates existing simulation methods without improving accuracy
• You are creating datasets as an end in themselves, without a prediction method
• Your AI model offers speed gains but no accuracy advantage over non-AI state of the art
• You are based at a for-profit institution

Core Question

Can AI help us make predictions for matter in the regimes that actually determine technological function: defects, interfaces, disorder, and dynamics at multiple length and time scales?

Description

Computational practice in chemistry, materials science, and biology has long emphasized idealized systems—because non-idealities and multi-scale phenomena are expensive to compute and hard to represent. Yet the functional properties of technologies we depend on (catalysts, semiconductors, drugs, batteries) are controlled by these very non-idealities operating at multiple scales.: a grain boundary, a surface defect, a disordered polymer chain. We believe the necessary ingredients to address this—AI architectures with physical symmetry, large simulated datasets, and sufficient compute—are now in place.

This RFP seeks proposals for AI-driven methods that can reliably predict properties governed by defects, interfaces, disorder, and multi-scale dynamics—including methods that bridge length and time scales that conventional simulation cannot connect—for any technologically relevant material or molecule.

We particularly encourage submissions from ongoing efforts where additional resources would enable a step-change in ambition. Teams with existing momentum—preliminary data, assembled collaborations between computationalists and experimentalists, validated workflows—are especially welcome. 

Successful projects may be considered for follow-on funding, and funded teams will be invited to a convening of researchers and industry leaders in this space (anticipated Summer 2026).

Funding Tracks

Applicants can apply for one of two funding track:

Track I: Up to $100,000 (Up to 12 months)

The goal is to produce preliminary evidence that a proposed AI-driven approach can credibly address a substantial, real-world materials or molecular prediction problem.

Track I is appropriate if you:

• Have a promising method or architecture not yet validated against experimental data in your target domain
• Need to generate or curate a critical dataset to demonstrate feasibility
• Want to establish that your approach offers meaningful accuracy gains over conventional methods before scaling up

Track II: $100,000 – $500,000 (12–18 months)

Track II funds the development of substantial AI-driven tools, particularly at a level useful to experimentalists. Track II is appropriate when feasibility has already been established through prior work, published results, or existing infrastructure.

Track II is appropriate if you:

• Have existing models, datasets, or preliminary results demonstrating that your approach is feasible
• Are ready to develop and validate a tool for a specific, high-value prediction problem
• Can show meaningful progress and deliver software/tooling outputs within 12–18 months

What we will fund

We will fund projects that deliver:

• AI-driven methods predicting non-idealities and/or multi-scale phenomena with meaningful accuracy improvements over conventional simulation
• Designs that make methods useful in experimental labs, not just competitive on curated benchmarks
• Well-mapped model failure modes

All technologically relevant materials and molecules (including biomolecules) are in scope.

What we will not fund (generally)

• "Interesting architecture" work without a concrete path to accuracy improvements over conventional computational methods
• Dataset creation as an end in itself
• AI models that offer speed advantages but no accuracy advantage compared to non-AI state of the art
• Work on materials or molecules with no clear technological relevance

Review Criteria

Submissions will be assessed by internal and external reviewers with expertise relevant to your proposal topic. During the assessment process, we may reach out to applicants to clarify portions of their proposal. Assessment criteria include:

• Readiness and momentum: Does the team have existing work, data, or infrastructure that positions them for rapid progress?
• Magnitude of improvement: How large and well-justified is the expected accuracy improvement over conventional simulation methods?
• Scale-bridging ambition: For multi-scale proposals, does the approach credibly connect regimes that are currently treated in isolation?
• Experimental relevance: How likely are model outputs to directly inform experimental work?
• Potential for field impact: Does the project have the potential to generate compelling evidence of what's possible in AI-driven matter modelling?

How to Apply

Documents should be submitted through the SurveyMonkey Apply portal. Below is a list of the items you will need to submit: 

1. Project Information 
   Name, email and title of the person submitting the proposal.
   Institution/Recipient of Fund. 
   Choice of Funding Track (I or II)
   Total Project Budget (USD)
   Project Start Date (MM/YY)
   Project End Date (MM/YY)

2. Team Composition and CVs
   
   Please upload the CV of the lead PI (mandatory) and other team members (optional) (Must be uploaded as a PDF.)

3. Project Abstract
   
   Please provide a short abstract of your project written for a non-specialist audience (< 200 words).

4. Bigger Ambition

   A brief sketch of where this work could lead in 5 years if this project succeeds and significant follow-on funding (upto $10M) were available. (< 150 words).
   
   
5. Size of follow-up funding
   
   If your experiments through this award were to be successful, what scale of funding would you seek next to drive towards the aforementioned bigger ambition?
   
   
6. Measures of Success
   
   Please indicate (< 100 words) 3-5 measures that we should use to gauge the success of the project when it concludes.
   
   

7. Project Narrative (pdf upload; up to 5 pages, excluding references) 
   Please describe your proposed project in a pdf, and address the questions listed below. Proposals must not exceed 5 pages, including figures but excluding citations.

   • Problem and Opportunity: Which specific non-ideality or multi-scale problem are you solving, and where is the technological relevance? Why do conventional methods underperform? Why is AI expected to help, and why now?

   • Approach and Methods: Describe the AI approach being developed and why it will yield accuracy improvements. Describe your data plan (what data you will use or create, and why it's sufficient) and your validation plan (what tests, what go/no-go thresholds). What are the risks, and how will you address them?

   • Environmental Scan: How does your project fit in with the existing work in this field? What are the new and distinguishing features of this approach?

   • Existing Capability & Momentum: Describe your team and what you have already built or demonstrated. Include preliminary data, or results if available.

   • Workplan: Brief work plan with milestones.

   • Budget: Provide a high-level budget breakdown. This is not a binding budget — detailed budgets will be requested from selected finalists. We want to understand how you plan to allocate resources at a rough level.

8. Is there anything else you'd like us to know? (optional)

For example: links to demo videos, slides, relevant press coverage, prior collaborations with industry, suggestions or context that doesn't fit elsewhere (optional).