(Illustration by Harry Campbell)
In 2003, the Institute of Medicine (now the National Academy of Medicine) declared that medical research was failing patients. Decades of brilliant basic science sat stranded in academic journals while people suffered from treatable diseases. The problem wasn’t a lack of discovery; it was a lack of translation. The relevant scientific insights existed, but the infrastructure to move them from the lab to bedside did not.
The solution was translational medicine, a new discipline supported by an ecosystem of specialized infrastructure—clinical-trial-design firms, regulatory navigators, data-management systems, and manufacturing expertise. Researchers didn’t build this infrastructure themselves, but their access to it transformed medical practice.
Education today faces a strikingly similar crisis. We have more knowledge about how people learn than at any point in human history. Cognitive science, neuroscience, social and affective science, and the learning sciences—the empirical study of how people learn—have produced robust insights that could meaningfully improve outcomes for students. Federal and philanthropic investment in education R&D is growing, and AI is creating new possibilities. But the disparity between medical and education funding in this area is stark: The National Institutes of Health alone receives approximately $47 billion annually for biomedical research—more than 50 times what the federal government invests ineducation R&D through the Institute of Education Sciences. As a share of sector spending, biomedical research receives roughly 6 percent of healthcare expenditures, while education R&D receives less than 0.1 percent of the $800 billion spent annually on K-12 education. Less than 0.2 percent of all federal R&D funding goes to the US Department of Education. (These figures reflect pre-2025 federal data; reporting has become uneven since, but the directional disparity has widened, not narrowed.)
Such disparities help to explain why the gap between what we know and what reaches classrooms remains vast. Promising interventions stall in pilot programs. Edtech products launch without evidence of efficacy. Researchers struggle to translate findings into usable tools. The problem isn’t discovery—it’s delivery.
Education needs its own engineering discipline: a systematic approach to moving innovations from insight to impact. I call it impact engineering. It is the discipline of applying scientific and engineering rigor to move educational innovations from discovery to delivery—ethically, efficiently, and at scale. Where learning engineering focuses on designing better learning experiences, impact engineering builds the full pipeline that gets those experiences to students at scale.
Learning Engineering Is Not Enough
For nearly a decade, I’ve advocated for applying engineering rigor to education. In 2016, I argued at a White House symposium that we needed “learning engineering” to close the loop between research and practice. In subsequent publications and alongside parallel efforts, I helped develop frameworks for how learning scientists, educators, and developers could work together to design effective learning experiences.
Learning engineering has since gained traction. The Institute of Electrical and Electronics Engineers (IEEE) established the International Consortium for Innovation and Collaboration in Learning Engineering (ICICLE) to define common standards for the field of learning engineering—shared methods, data formats, and design principles for practitioners. Universities launched graduate programs for learning engineering. The field now has conferences, a community-maintained tool kit of methods and shared resources for learning-engineering practitioners, and a growing community that applies learning science using human-centered design and data-informed decision-making to improve learner outcomes.
Consider, for example, the delivery of an adaptive tutoring system—one that adjusts in real time to what each student gets right, gets wrong, and needs next. Learning scientists know that revisiting material across spaced intervals helps students remember it long-term; human-centered designers work with teachers to ensure that the system fits classroom workflows; data scientists build models that track student responses and trigger personalized review sequences. The result is a system that continuously improves based on evidence—not a static curriculum. Such progress matters. Learning engineering addresses a critical gap between research evidence and designing effective learning experiences.
After five years leading R&D infrastructure at the Advanced Education Research and Development Fund (AERDF), I’ve come to see that learning engineering—while essential—addresses only part of the problem. Designing effective learning experiences is necessary, but it’s not sufficient for getting those experiences to students at scale, especially students who have been historically underserved. The bottleneck concerns the entire translation pathway, from discovery through demonstration to delivery.
Consider what happens when a researcher discovers something promising about how students learn, or when an entrepreneur builds an edtech product that shows encouraging initial results. To convert that innovation into impact, they need to navigate Institutional Review Board (IRB) protocols for research with minors. They need access to a secure data infrastructure that complies with federal privacy protections and state privacy laws. They need data-sharing agreements with districts. They need scientific review of their methodology. They need evaluation designs that can demonstrate efficacy. And they need reliable pathways to scale.
In biomedicine, such infrastructure exists and has transformed how medicine reaches patients who need it. The National Center for Advancing Translational Sciences (NCATS), established in 2012, coordinates a national network of more than 60 academic medical centers that move lab discoveries to treatments that patients can actually receive. Specialized firms handle clinical-trial management, regulatory navigation, data compliance, and manufacturing scale-up. The NCATS Trial Innovation Network has accelerated dozens of clinical trials across therapeutic areas, including rapid COVID-19 trials. Every dollar invested in National Institutes of Health (NIH) research generates $2.56 in economic activity because the infrastructure exists to translate discoveries into treatments that reach patients.
In education, this infrastructure barely exists. There is no equivalent to NCATS, no network of translation centers, no ecosystem of specialized firms supporting education R&D at scale. The expertise is sparse, expensive, and fragmented. Every R&D team has to craft its own privacy compliance de novo. Every edtech startup struggles through district procurement. Every funder watches promising grantees stall because they lack the engineering capacity to execute.
The biomedical example points the way to a better system. What enabled that field’s success was not just good science, but the deliberate creation of a translational discipline with dedicated infrastructure, professional training, and shared resources. Education needs the same. It needs impact engineering.
Where learning engineering focuses on designing effective learning experiences, impact engineering encompasses the full translation pathway: identifying problems worth solving, codesigning solutions with educators and students, building and iterating prototypes with technical excellence, demonstrating impact with scientific rigor, and delivering innovations into diverse real-world contexts. Crucially, impact engineering requires what I refer to as “full-stack” infrastructure—expertise spanning ethics, privacy, data architecture, scientific review, evaluation design, impact analysis, and policy navigation—integrated into every stage of R&D.
A Call to Funders and Builders
At AERDF, we built this infrastructure for programs tackling some of education’s most persistent challenges: EF+Math (improving mathematics learning by strengthening executive function), Assessment for Good (rebuilding assessment to measure what students bring rather than what they lack, and to reflect their identities rather than penalize them), and Reading Reimagined (developing foundational literacy skills).
We learned that the infrastructure mattered as much as the innovations themselves. Teams with access to ethical-review support, secure data systems, privacy expertise, and scientific guidance could move faster and more confidently. Teams without it stalled—not because their ideas weren’t good, but because they lacked the support they needed.
We also learned that this infrastructure is highly intersectional: A single challenge might require expertise in data science, privacy law, cloud architecture, IRB protocols, and educator engagement—expertise that rarely exists in one organization. The solution isn’t to train every researcher in every domain, but to build accessible infrastructure that teams can draw upon.
If we’re serious about improving educational outcomes—especially for students who have been least served by the current system—we need to invest in the engineering infrastructure that makes breakthrough R&D possible. Infrastructure is not overhead. It’s the backbone that determines whether investments translate to impact. Every dollar spent on promising research that stalls due to preventable obstacles—a privacy violation, a failed IRB submission, a district partnership that collapses—is a dollar that fails students.
The efficiency gains from building shared infrastructure that multiple R&D efforts can access, rather than expecting every team to build their own, are substantial. The quality gains are even greater.
The vision that motivated my advocacy for learning engineering a decade ago—that we can and must apply rigorous, evidence-based approaches to education—remains urgent. Learning engineering has made important progress on that vision, particularly in how we design learning experiences.
But design is not enough. We need the full translation pathway. We need infrastructure that enables researchers, educators, and developers to focus on innovation while the engineering backbone handles the complexity of ethics, privacy, data, and scale.
Impact engineering must also become a recognized discipline—with training programs, professional standards, and career pathways—just as biomedicine built translational science into a discipline over the past two decades. Translational science didn’t just move discoveries to patients faster. It created an entirely new field, with its own methods, institutions, and career pathways. It changed what was possible. Impact engineering can do the same for education.
Read more stories by Melina R. Uncapher.