Meerkat Privacy: Web Privacy Extension

Overview

Meerkat Privacy is a collaborative project aimed at making website privacy policies understandable for everyday users. As Web Developer, I focused on three areas: integrating the OpenAI (ChatGPT) API with our crawling pipeline, building the marketing/onboarding website, and designing the end-to-end UX flow (including a preference panel and risk-scoring UI). The result is a practical browser extension that surfaces plain-language takeaways right inside the browsing experience.

‍

Key Feature

Problem / Challenge

• Privacy policies are long, legalistic, and time-consuming to read.
• Users lack quick, trustworthy summaries about data collection, sharing, and risks.
• Insights should appear in context (in the browser), not on a separate, hard-to-find page.

Goals & Approach

Goals

• Deliver concise, readable summaries of policies.
• Provide risk visibility (e.g., data sharing, tracking, retention).
• Offer a simple onboarding so users know how/when summaries appear.

Approach

• Use web crawlers to fetch relevant policy sections (collection, usage, third-party sharing).
• Send content to OpenAI’s API with structured prompts to produce sectioned, scannable summaries.
• Present results in a browser extension popup with a preference panel and risk scoring.
• Support the experience with a website for feature overview and onboarding.

Team & My Role

• Collaboration: Worked in a cross-functional team (engineers + design).
• My Role: Full-Stack Developer / UX Contributor
  • API Integration: Implemented the ChatGPT API flow and wired it to our crawlers; handled chunking, retry, and error states.
  • Frontend Website: Built the project website to showcase features and provide user onboarding.
  • UX Design: Designed the end-to-end UX flow, including preference panel (controls for depth/sections) and a scoring system that surfaces risk at a glance.

Development Process

Tech Stack

• Chrome Extension APIs
• Web Crawlers / scraping pipeline
• OpenAI (ChatGPT) API

Key Implementation Details (my contributions)

• Policy ingestion: Normalized and chunked long documents before LLM calls to respect token limits and keep latency acceptable.
• Structured prompting: Standardized outputs into labeled sections (Data Collected / Usage / Sharing / Retention / User Rights) to avoid vague summaries.
• Risk model & UI: Defined a simple scoring rubric and mapped it to visual indicators; integrated a preference panel so users can tune detail level.
• Website & onboarding: Built a responsive site explaining the extension’s value, install steps, and how summaries appear during browsing.

Key Features

• In-browser summaries: One-click popup with plain-language insights.
• Risk scoring: Clear visual signal for potential concerns.
• Preference panel: Users choose summary depth/sections.
• Auto-fetching: Crawlers locate and retrieve policy content without manual copy-paste.
• Onboarding site: Guides installation and showcases use cases.

Outcome / Results

• Delivered a working MVP extension with a coherent UX and a supporting onboarding website.
• Transformed dense legal text into scannable, structured insights, improving comprehension and decision-making.
• Established a modular pipeline (crawl → summarize → display) that’s ready to scale and iterate.

Learnings & Next Steps

• Prompt/format stability is key to trustworthy summaries; future work: evaluation sets + guardrails.
• Add browser coverage (beyond Chrome) and caching to reduce repeated API calls.
• Explore user alerts (e.g., flag high-risk clauses) and export/share options for compliance teams.

‍