NSF-NIH Smart Health (SCH) NSF 25-542: $20M Annually for Biomedical AI With a September 10 Target Date

The Smart Health and Biomedical Research in the Era of Artificial Intelligence and Advanced Data Science program — universally known as SCH inside the agencies that run it — sits in an unusual administrative position. It is an NSF solicitation. It is also an NIH solicitation. It is jointly written, jointly reviewed, and jointly funded by both agencies, with more than twenty NIH institutes and centers participating alongside three NSF directorates. The current iteration, NSF 25-542, opens for full proposals on a twice-yearly target schedule. The next opportunity in the cycle is September 10, 2026.

For researchers working at the intersection of artificial intelligence, advanced data science, and biomedical or public health questions, SCH is one of the few federal mechanisms designed from the ground up to fund the kind of cross-disciplinary work that neither agency alone can comfortably support. Understanding how SCH actually selects winning proposals — as opposed to how its abstract describes the program — is the difference between submitting work that scores in the top quartile and submitting work that gets returned without review.

Tracking biomedical AI funding? Browse our Healthcare AI Grants Hub for federal and foundation funding across digital health, biomedical informatics, and clinical AI.

The Money and the Math

The program operates with a $15 million to $20 million annual budget, awarding approximately 10 to 16 projects per year. Maximum award size is $300,000 per year for up to four years, with a total project ceiling of $1.2 million. Awards are issued as Standard Grants, Continuing Grants, or Cooperative Agreements, depending on the NSF or NIH component leading the funding decision.

The implicit success rate at this funding level is meaningful. Twenty SCH directorates and NIH institutes typically draw a few hundred proposals across each cycle, which means funding rates fluctuate between 5% and 12% depending on the year and the cycle. That is below the average NSF rate but in line with NIH R01 success in computationally intensive study sections — and crucially, the work that gets funded tends to be work that would not have been competitive in any single-agency mechanism.

The twice-yearly target dates — the first Thursday in February and the second Thursday in September — are target dates, not hard deadlines. Proposals can technically be submitted at any point in the cycle, but program officers cluster reviews around the target dates to allow joint panels to assemble. Submitting a proposal off-cycle is operationally equivalent to submitting it for the next target date.

The Six Research Themes

SCH organizes its priorities into six thematic areas. Proposals are not required to land squarely in one theme, but proposals that cannot identify their primary theme rarely fare well in panel discussion. The themes are:

Fairness and Trustworthiness in AI/ML. Methodological work on bias, interpretability, and reliability in models that touch clinical or public health decision-making. This theme has seen growing investment because the failure modes of clinical AI are increasingly the failure modes that program officers and review panels worry about. Proposals here need a credible biomedical use case, not just a methods contribution.

Transformative Analytics in Biomedical and Behavioral Research. Statistical, machine learning, and causal inference methods applied to biomedical questions where the analytic challenge is the bottleneck. The strongest proposals identify a specific biological or behavioral question where the existing analytic toolkit is genuinely insufficient.

Next Generation Multimodal and Reconfigurable Sensing Systems. Wearables, ambient sensors, and integrated sensing platforms that produce streams of data requiring new analytic approaches. This theme rewards hardware-software-analytics integration, not just one of the three.

Cyber-Physical Systems. Closed-loop systems that sense, decide, and act in clinical or public health contexts. Continuous glucose monitoring with automated insulin delivery is the canonical example. Newer applications include closed-loop neurostimulation and adaptive ventilator control.

Robotics. Robotic systems in clinical, rehabilitative, or home health contexts. Proposals here need to identify the specific clinical task that the robotic system addresses and the patient population that benefits — not just propose a robotic platform with general medical applications.

Biomedical Image Interpretation. AI for medical imaging, with emphasis on going beyond classification tasks into more complex interpretation, prediction, or generation work. Proposals that essentially propose to train a deep learning model on a labeled imaging dataset will not score well — the field is past that point.

What the Review Process Actually Looks Like

SCH proposals are reviewed under standard NSF merit review criteria — Intellectual Merit and Broader Impacts — but with joint panels that include both NSF and NIH program officers, and reviewers drawn from both communities. The practical implication is that a proposal must speak credibly to both audiences.

NSF reviewers will weight the methodological novelty heavily. They want to see that the proposed work advances knowledge in AI, data science, sensing, or robotics — not just that AI is being applied to a biomedical problem. NIH reviewers will weight the clinical or biological significance heavily. They want to see that the proposed work addresses a question that, if answered, would change biomedical understanding or clinical practice.

Proposals that lean too hard in either direction get filleted in panel. A proposal that proposes a beautiful new method but cannot articulate why any clinician would care about the answer will be sent back. A proposal that addresses an important clinical question but proposes off-the-shelf methods will be sent back. Strong SCH proposals find the seam where methodological advance and biomedical significance both happen because of each other.

PIs may participate in no more than two SCH proposals per annual deadline across all roles — PI, co-PI, or senior personnel. This limit exists because the program is small relative to the community it serves, and the limit is enforced strictly. PIs who appear on more than two proposals will see their excess proposals returned without review.

The Cross-Disciplinary Team Requirement

SCH proposals require demonstrated cross-disciplinary teams. A proposal led by a computer scientist with a clinician listed as a consultant will not score well. The program is funding genuine collaboration between methodologists and domain experts, and the proposal must read like both communities co-authored it.

The strongest team structures pair a methodologist PI with a clinician or public health researcher co-PI, with shared decision-making authority over the research direction. Letters of collaboration matter, but the proposal narrative matters more. Reviewers can tell within the first three pages whether the methodologist understands the clinical context and whether the clinician understands what the methods can and cannot deliver.

The Broader Impacts section is where many SCH proposals lose points unnecessarily. The program expects that a $1.2 million investment in biomedical AI research produces benefits beyond the published papers. Proposals that articulate specific pathways to clinical translation, to open-source software release, to training the next generation of cross-disciplinary researchers, or to community engagement around the technology will score better than proposals that treat Broader Impacts as a perfunctory closing section.

How to Use the Time Between Now and September

The September 10 target date is sixteen weeks out. Proposals that begin development now have time to do the work that distinguishes funded proposals from declined ones: building a credible preliminary data section, securing concrete letters of collaboration from clinical or public health partners, refining the methodological approach so that the technical contribution is clearly articulated.

Three specific moves matter.

First, contact a program officer. SCH program officers actively encourage early engagement with applicants. A fifteen-minute call before drafting begins can save weeks of misdirected effort. Program officers can tell you whether your proposed direction is competitive in the current portfolio, which themes are over- or undersubscribed, and whether your approach should be positioned as primarily NSF-funded or primarily NIH-funded.

Second, identify and engage your domain partner now. Clinician collaborators on cross-disciplinary AI proposals are oversubscribed at most academic medical centers, and the strongest partnerships are built through prior shared work. If you do not have a clinical or public health collaborator with whom you have published or run a pilot, the September deadline is probably not the right target. The February 2027 deadline gives more realistic time to build that relationship.

Third, read funded SCH abstracts on the NSF awards database. SCH funding decisions are surprisingly readable through their abstracts. Patterns emerge — the kind of methodological novelty that gets funded, the kind of clinical framing that survives panel review, the kind of cross-disciplinary team structures that the program is actually willing to support. Twenty hours spent reading recent awards will reshape a proposal more than fifty hours spent polishing prose.

SCH is one of the few federal mechanisms where a thoughtful, well-positioned proposal from a credible cross-disciplinary team has a realistic shot at competing for transformative funding. The selection process rewards specificity and penalizes hand-waving. Researchers who match their proposed work to what the program actually funds — not what they wish it funded — have a real path here. Tools like Granted can help research teams identify the broader landscape of federal and foundation funding that complements SCH support across the AI-in-health continuum, from foundational methodology grants to clinical translation funding to commercialization pathways.