Showcase: SafeDraft AI - A CISSP-designed Privacy Firewall for LLMs (Built with Next.js & Vercel)

[willbiu]

Project Overview

As a Data Security Executive (CISSP), I’ve seen too many professionals hesitate to use LLMs due to PII leakage risks. I built SafeDraft AI to solve this "Privacy-Utility Dilemma."

It’s a specialized de-identification tool that uses Reversible Masking—allowing users to scrub sensitive data locally before feeding it to AI, then map it back using a secure local key.

Tech Stack

• Framework: Next.js 14+ (App Router)
• Deployment: Vercel (Leveraging Edge Functions for low-latency scrubbing)
• Styling: Tailwind CSS
• Compliance: Strict Zero-Retention architecture

Key Features

• Reversible De-identification: Unlike simple redaction, our placeholders maintain AI context while protecting privacy.
• Privacy-First: All sensitive mapping happens in the browser/edge; no data is stored on our servers.
• Freemium SaaS Model: Structured with anonymous, registered, and pro tiers to support professional workflows.

Links

• Live Site: https://safedraftai.com/
• Product Page: Indie Hackers - SafeDraft AI

I'd love to hear the community's thoughts on our de-identification precision and the UI/UX! 🚀

[willbiu]

Headline: Big step for SafeDraft AI: Submitted to the official Next.js Showcase! 🚀

Content: > Just shared our tech architecture and security logic with the Vercel/Next.js community on GitHub!

Even though our core engine remains proprietary for security reasons, I've detailed our Zero-Retention architecture and how we leverage Vercel Edge Functions to ensure low-latency de-identification.

Check out the discussion and our tech stack here: [https://github.com//discussions/91596]

If you're a fellow dev, I'd love your upvote or a comment on our masking logic!