National Institutes of Health
Office of Dietary Supplements · Supplements, Facts First · 2026
The technology that won an NIH award.
ConductScience built a scan-first, multi-modal system that turns static NIH supplement fact sheets into a personalized safety tool — and it was named a Phase 1 winner of the NIH “Supplements, Facts First” challenge. Here is what we built.
The recognition
One of eight teams chosen nationwide.
The NIH Office of Dietary Supplements named Resin Health a Phase 1 winner of Supplements, Facts First: A Digital Adventure for Every Age — a challenge to turn authoritative but static supplement fact sheets into experiences people actually use. The recognition advances the project into Phase 2, the prototype round, where up to five teams move forward.
0
Phase 1 winners nationwide
0
Modalities built (two required)
0+
NIH ODS fact sheets connected
0
Explainer videos in the planned library
The challenge
Turn authoritative fact sheets into something people actually use.
The NIH Office of Dietary Supplements publishes the most comprehensive supplement fact sheets available, covering more than 100 nutrients and botanicals. The science is thorough. The problem is that it stays on a government website, disconnected from the bottles people buy and the medications they take.
The challenge asked teams to convert those static fact sheets into engaging digital experiences that reach diverse populations across every age, proving the concept across at least two modalities. We built three, and connected them into a single journey that starts with a physical scan.
What we built
Scan, Check, Decide — one system, three modalities.
A single scan drives a visual dashboard, an ODS-grounded chatbot, and short explainer videos. Here is the technology underneath it.
One tap on a bottle produces a personalized safety dashboard: each ingredient mapped to its ODS fact sheet, cross-checked against the user’s medications, and returned as plain-language flags with concrete next steps — never abstract risk scores.
01
A scan engine that reads any bottle
Barcode detection runs against a cached index of NIH Dietary Supplement Label Database (DSLD) products — instant, and fully offline. When a barcode is unknown (store brands, imports, new products), the app falls back to ConductVision OCR, our computer-vision engine, which reads the Supplement Facts panel directly from the camera and extracts every ingredient, dose, and daily value.
02
A personalized interaction engine
Each extracted ingredient is mapped to its ODS fact sheet and cross-checked against the user’s own medication profile. The result is a red / amber / green safety dashboard with concrete, plain-language next steps rather than abstract risk scores — the output of a rules engine built on authoritative NIH interaction data.
03
A retrieval-grounded chatbot
From any flag, users can ask a question in plain language. The chatbot uses retrieval-augmented generation over structured ODS content: it embeds the question, retrieves the relevant fact-sheet passages, and answers with a citation back to the source. It is deliberately constrained — when a question falls outside ODS coverage, it says so and points to a provider rather than inventing an answer.
04
An AI explainer-video pipeline
Every fact sheet becomes a 60–90 second video through an automated pipeline: an LLM generates a demographic-specific script, neural text-to-speech renders a warm narration, and ingredient, dosage, and safety frames are composited with WCAG-AA captions. The planned library is roughly 300 videos, pre-generated and served from a CDN, with 9:16 exports for social distribution.
05
An offline-first PWA architecture
The core principle is to pre-structure all data and eliminate runtime cloud dependencies. Every NIH source is downloaded, processed into structured files, and bundled for local or CDN delivery. The app is a Progressive Web App built on Next.js and deployed on Vercel — no app store, no download. A service worker caches supplement data and fact sheets so the scan-and-dashboard experience works with no network at all.
06
Accessibility and equity, engineered in
The interface ships a high-contrast mode with 16px minimum text and 44px touch targets for older adults, plain-language summaries at a sixth-grade reading level, and color-coded indicators that reduce reliance on text. Removing the app-store barrier is itself the equity intervention — the tool reaches people on older phones, limited data, and low digital literacy.
07
Grounded entirely in authoritative NIH data
Nothing is invented. Every answer traces to the NIH Office of Dietary Supplements fact sheets, the Dietary Supplement Label Database, and DailyMed interaction data. Refusing to generate beyond that source is a feature, not a limit — it is what keeps the science trustworthy.
How we designed it
Engineering in service of the people who use it.
The technology is only half the work. The other half is who it is for.
Behavior-change science
Every choice maps to the COM-B model. The scan is Opportunity, demographic-adapted content builds Capability, and personalized flags sustain Motivation.
Three audiences, each justified
Older adults 65+ with polypharmacy, prenatal and postpartum mothers, and community pharmacists as a distribution amplifier — each with its own content, defaults, and safety thresholds.
Community co-design
We partnered with Link Health, embedded in safety-net clinics across Boston and Houston. Their Certified Patient Navigators shaped the concept following community-based participatory research principles.
What’s next
Phase 2 is underway.
We’re building the working prototype and running user testing with community partners in Boston and Houston, through 2027. See the product, or talk to us about partnering.
