Building a Science-First Ingredient Interpreter
Overview:
In collaboration with AI software engineer Keon Sadatian, I’m designing and prototyping a tool that translates skincare ingredient lists into science-based, practical explanations.
Ingredient data is complex, inconsistent, and often stripped of meaningful context for the user. This ongoing project explores how to design a content system and experience layer that interprets ingredients through scientific use cases, risk levels, and user context, while laying the foundation for scalable AI integration. It’s a hybrid of UX content strategy, systems thinking, and early GenAI readiness.
What’s New
Iterating on modular UX patterns in Figma for ingredient outputs and contextual warnings
Refining tagging structure and metadata models for future AI parsing and filtering
Exploring content fallback strategies for ambiguous or high-risk ingredient responses
Structuring content logic and output formats to align with future LLM-based interpretation
View the deeper dive into the newest system updates here →
The Core Interaction:
Paste a product’s ingredient list → Get clear, contextual, evidence-based explanations.
Too often, ingredient checkers leave users with vague warnings, alarmist scores, or trend-chasing jargon. There’s a disconnect between what people want to understand and how they’re being spoken to.
We’re Building This Tool to…
Explain each ingredient’s function using readable, accurate language
Provide ingredient context (e.g., concentration, formulation type)
Avoid moralizing or fear-based terminology
Offer credible references for deeper learning
Summarize overall product composition (e.g., “4 humectants, 2 emollients”)
We don’t follow “clean beauty” systems like EWG. Our approach is rooted in evidence—not anxiety.
What We’re Exploring:
Ingredient Data Sources
We’re gathering trusted data from:
These include INCI Decoder,
CosIng (EU database)
UX Patterns
Ingredient-by-ingredient breakdowns
Summary view highlighting overall product structure
Reference links to support transparency and trust
AI Capabilities
GPT-based models decode INCI terminology
Ingredients classified by role (e.g., humectant, preservative)
Use cases included (e.g., hydration, exfoliation)
Contextual flags added (e.g., “may irritate above 1%”)
Prompts engineered to prioritize calm, accurate language
Example Output
“Phenoxyethanol: A preservative that stops bacteria in water-based formulas. Usually well-tolerated, but may cause irritation at high levels (>1%). Often found in cleansers and moisturizers.”
See: PubChem, CIR Report
Collaboration and Role:
Keon and I are collaborating closely across multiple parts of the project:
Data sourcing and structure: I lead research on ingredient sources and interpretation. Keon develops efficient methods to organize and retrieve that data.
AI prompt engineering: I write and refine GPT prompts for clarity, neutrality, and scientific accuracy. Keon tests model responses and ensures backend performance.
System architecture and design: We co-develop how content is presented, refine how responses are surfaced, and explore UI controls for transparency and trust.
From idea to prototype: I lead UX flow and interface design. Keon builds frontend/backend prototypes for iterative testing.
Phases:
Phase 1: Testing + Iteration (Current Phase)
Integrate with LLM or retrieval-augmented model
Conduct usability testing on language clarity and risk phrasing
Align outputs with accessibility and enterprise design systems
Phase 2: Testing + Iteration
Begin user testing with real ingredient lists
Test upload and paste workflows
Build a trust-feedback loop to guide future improvements
Future Plans
Layer in user profiles (e.g., acne-prone, sensitive skin, rosacea)
Enable “smart” questions (e.g., “Is this pregnancy safe?”)
Develop a Chrome extension for ingredient popovers while brows