Fragmented public data
Recall notices are published by multiple agencies in different formats, making them hard to unify and hard for consumers to monitor consistently.
The Problem
Recall information exists, but shoppers rarely see it where they make decisions.
No in-context warning
Online grocery platforms usually do not surface recall information directly on product pages, even though that is where the buying decision happens.
Messy product matching
Retail product names do not perfectly match recall wording, so a useful solution needs more than exact string comparison.
The Solution
BiteWise turns recall records into a real-time product safety signal.
I built BiteWise as a local full-stack prototype that connects public recall data to the online shopping experience. The system collects current recall records, stores them in a normalized SQLite database, exposes a matching API, and lets a browser extension flag relevant products directly on Instacart pages.
The matching logic combines fuzzy similarity with practical guardrails such as company-name downweighting, distinctive token overlap, size checks, and a 180-day freshness window to reduce false positives.
Product Demo
Three shopper-facing outcomes inside the browser.
High-confidence alert
Red banner
A strong match between the product and an active recall. The shopper gets an immediate warning, a recall date, an age indicator, and a direct link to the official recall notice.
Brand-level caution
Yellow banner
A meaningful but weaker match. This warns the shopper that the brand is linked to a recent recall even if the exact product evidence is less certain.
No active match
Green banner
When BiteWise does not find a relevant active recall, the extension confirms that no active recall match was found for the current product.
How It Works
A straightforward pipeline from public recalls to in-browser alerts.
Collect recall data
FDA recall records are scraped from the web, while FSIS records are pulled through a public API.
Normalize and store
Recall data is cleaned and loaded into SQLite with separate recall and product tables.
Match products
The Flask API compares Instacart products against recent recall records using fuzzy scoring and rule-based safeguards.
Alert in context
The Chrome extension injects a clear status banner directly into the shopping page.
Technologies
Built as a compact full-stack system.
Data ingestion
Python, Requests, BeautifulSoup
Storage
SQLite with a normalized recall and product schema
API
Flask and Flask-CORS
Matching
RapidFuzz plus custom heuristics and guardrails
Frontend delivery
Chrome Extension (Manifest V3) and vanilla JavaScript
Quality
Unit tests for matcher behavior and API responses
How It Can Scale
The prototype is local today, but the architecture can grow.
Broader retailer support
Extend the browser layer to other grocery and e-commerce platforms with platform-specific selectors.
Automated refresh jobs
Move scraping and ingestion to scheduled background jobs so recall data stays continuously updated.
Production deployment
Swap SQLite for a managed database, host the API, and precompute/cache recall candidates for faster lookups.