Case Study · Development
SaaS nutrition label generation for Canadian cottage food producers. Regulatory compliance at a price that makes sense.
Cottage food producers in Canada are required by law to put accurate, regulatory-format nutrition facts labels on packaged goods. The compliant path — hiring a registered dietitian or sending samples to a commercial lab — can run hundreds of dollars per recipe. And it doesn't scale when recipes change.
Existing tools were either generic nutrition calculators (no regulatory output, no audit trail) or enterprise software built for industrial food manufacturers with budgets to match. Nobody had built for the cottage producer.
A production SaaS that takes a recipe, maps each ingredient to authoritative government nutrition databases, calculates per-serving nutritional data to regulatory rounding rules, and generates print-ready Nutrition Facts label PDFs in Canadian (bilingual) and US formats.
How it's built
The architecture is organised around a single constraint: a calculation isn't trustworthy unless you can show where it came from. That shaped every layer, from the database query path to the PDF on the page.
Health Canada's Canadian Nutrient File and the USDA Standard Reference, normalized into a unified query layer despite meaningfully different data structures.
Every ingredient lookup is logged against its source database row, version, and timestamp — the calculation is traceable, not just stored.
Pixel-precise regulatory label layouts using ReportLab, a programmatic layout engine rather than a template filler.
Paste any recipe from the web and Claude AI parses ingredients and maps them to database entries automatically.
A three-tier subscription model — Free, Recipe Pro, Label Pro — with Stripe Checkout and webhook-based state management across all lifecycle events.
Persistent recipe management with per-user data isolation and label history.
Python / FastAPI · React · MySQL · AWS Lightsail · Docker · Stripe API · Claude API · ReportLab.
What this shows
The interesting technical challenge here wasn't the PDF generation or the Stripe integration — those are solved problems. It was designing a data system where every calculation is court-defensible. Every number on a Nutrifax label traces back to a specific government database row, with a timestamp. That's a different design constraint than most SaaS applications face, and it shaped every layer of the architecture.
The AI integration is also worth noting. Pasting a recipe URL and having ingredients automatically matched to database entries removes the single biggest friction point for non-technical users — and it works reliably because the matching logic is built around the database structure, not just free-text parsing.
Audit trails, traceable calculations, and data systems that hold up under scrutiny — from one practice, end to end.
