Lixus Agent — Skincare Intelligence Firewall

Autonomous 2026 Retail Scout. Lixus bridges the "Trust Gap" by cross-referencing Perfect Corp biological data with Gemini-powered fraud audits to secure your health and your wallet.

Comment

Lixus — Your Biological Skincare Firewall

Know your skin. Trust your products. Never buy something that conflicts with your biology again.

Lixus is an AI-powered mobile skincare intelligence platform. It uses Perfect Corp's clinical skin analysis API combined with Google Gemini to build a personalized Biological Safe State for each user — a real-time map of their skin vulnerabilities — and cross-references that map against the ingredient list of any product they're about to buy. The result is an instant, science-backed verdict: SAFE, CAUTION, or DANGER.

💡 Inspiration

In 2026, the global marketplace is a \$500B frontier of both innovation and deception. Consumers face a dual threat: Biological Mismatch and Retail Fraud. Inspired by the ancient gateway of Lixus, I built this "Biological Firewall" to ensure nothing touches your skin unless it is 100% compatible with your unique nature.

🛠 How We Built It

Lixus follows an Agentic Orchestrator architecture to move from data to decision in seconds:

  • Mobile Frontend: Built with React Native (Expo) featuring a "Liquid Glassmorphism" UI and ML Kit for real-time face detection.
  • Backend: A Node.js server utilizing Sharp.js for industrial-grade image preprocessing before hitting the Perfect Corp S2S API.
  • Intelligence Core: Gemini 2.5 Flash handles the "Reasoning Loop," while Open Beauty Facts provides the fallback database for barcode scanning.

The Security Formula

We calculate the compatibility score using a weighted non-linear equation:

$$S_{comp} = \sum_{i=1}^{14} (W_i \cdot C_i) - \Delta_{fraud}$$

The Variables:

  • ( S_{comp} ): The final Compatibility Score (0–100).
  • ( W_i ): The Severity Weight of the ( i )-th skin concern detected by Perfect Corp.
  • ( C_i ): The Conflict Factor (0 or 1) based on chemical ingredients identified by Gemini.
  • ( \Delta_{fraud} ): The Security Penalty applied if the Hudhud Scout detects suspicious retail indicators.

🧬 Perfect Corp Integration

Lixus uses the Perfect Corp YouCam Experience (YCE) S2S API v2.0 as the primary skin analysis engine. This is the backbone of the product; without accurate, clinical-grade skin measurement, downstream conflict detection would be meaningless.

Severity Classification Logic

The system classifies vulnerabilities based on the raw JSON output from the analysis task:

  • $\text{Score} \geq 90 \rightarrow$ Healthy (Strength)
  • $75 \leq \text{Score} < 90 \rightarrow$ Mild (Vulnerability)
  • $60 \leq \text{Score} < 75 \rightarrow$ Moderate (Vulnerability)
  • $\text{Score} < 60 \rightarrow$ Severe (Vulnerability)

🚧 Challenges Faced

The build wasn't without "Hackathon Nightmares":

  • The 429 Crisis: I hit the Gemini rate limit during testing, forcing the implementation of a Local Sentry fallback using regex-based ingredient checking.
  • The Physical Handshake: Getting the Samsung ADB link to work on a modern Mac required a deep dive into Android SDK 36 licenses and NDK 27 installations.
  • Emulator Lag: Heavy AI reasoning loops caused massive lag, solved by pivoting entirely to physical device testing for the final demo.

🎓 What I Learned

I learned that in 2026, AI is no longer just about "chatting"—it's about agency. Building an app that can investigate a URL and decide if a product is safe for a specific human's biology taught me the power of combining clinical sensor data with large-scale reasoning.

🚀 Future Vision

Scaling the Hudhud Match engine into a global affiliate network, allowing Lixus to suggest "100/100 compatible" products in real-time at the best price worldwide.

Built With

Share this project:

Updates