Silent Health: The Neuro-Audit Protocol

Clinical Report:
Clinical Report
Graphical Presenation
Hero Page

📖 The Story

Inspiration: From Mechanical Systems to Biological Ones

I am not a traditional software engineer. My background is in Mechanical Project Engineering—designing physical systems, analyzing tolerances, and ensuring structural integrity. In my field, we rely on rigorous audits to detect failure points before they become catastrophic.

Silent Health was born from applying this same engineering rigor to the human body. Just as a vibration sensor detects a misalignment in a turbine, I wanted to build a tool that could detect the subtle, "silent" micro-tremors and facial masking that signal the onset of Parkinson's Disease—long before they are visible to the naked eye.

The "Iron Man" Moment: Empowered by Google Antigravity

For years, the barrier to building complex, AI-driven medical software felt insurmountable for someone outside the "tech bubble." I knew what I wanted to build, but the how was trapped behind layers of complex infrastructure, cloud deployments, and steep learning curves.

Google Antigravity changed everything.

Using Google Antigravity along with the Gemini 3 models, I was able to bypass the traditional "learning cliff." The CLI didn't just write code; it acted as a senior architect, helping me structure the application, debug complex race conditions, and implement advanced features like real-time tensor processing in the browser.

Speed of Execution: 3 Days from Concept to Audit

The most staggering part of this journey is the velocity. This entire platform—including the real-time FaceMesh integration, the audio jitter analysis, the IndexedDB storage engine, and the complex risk scoring algorithms—was built and tested in just 3 days.

Gemini 3 Pro: I used the Pro model for the heavy lifting—planning the architecture, designing the "Audit Protocol," and ensuring the clinical logic was sound. Its reasoning capabilities allowed me to simulate edge cases without needing a QA team.
Gemini 3 Flash: For execution, the Flash model was my "infinite intern." It generated the React components, optimized the Tailwind CSS for that premium "medical-grade" aesthetic, and wrote the unit tests that ensure reliability.

What it Does

Silent Health is a zero-install, privacy-first web application that turns any laptop or phone into a neurological screening device.

Visual Audit: Uses MediaPipe FaceMesh (78 landmarks) to track "Hypomimia" (facial masking) and blink rates.
Vocal Audit: Analyzes voice stability to detect micro-tremors (Jitter/Shimmer) invisible to human hearing.
Longitudinal Tracking: Stores strictly local data to track disease progression over weeks or months.

How We Built It

Tech Stack: Next.js 15, React 19, Tailwind v4, Dexie.js (IndexedDB).
AI Engine: Google Gemini 3 (Pro & Flash) via Google Antigravity.
Privacy: All processing happens on-device. No video or audio is ever sent to a cloud server.

Challenges We Ran Into

The biggest challenge was handling the massive stream of real-time sensor data (30FPS video + 44.1kHz audio) without crashing the browser. Being a mechanical engineer, I approached this formatted as a "flow rate" problem. I worked with Gemini to implement an AtomicReading pattern—compressing data into lightweight snapshots—and used Web Workers to offload the processing, ensuring the UI remained buttery smooth.

Accomplishments That I'm Proud Of

I am most proud of the fact that this essentially "impossible" project for a single mechanical engineer became reality in a long weekend. It proves that with the right AI tools, domain expertise (engineering rigor) matters more than syntax memorization. We didn't just build an app; we built a Clinical Protocol.

What's Next for Silent Health

We plan to introduce "Passive Monitoring"—allowing the audit to run in the background during typical Zoom/Meet calls to gather data without requiring active patient sessions. We also aim to upgrade the "Risk Scoring" model with a custom-trained Gemini variant fine-tuned on anonymized clinical datasets.

🛠️ Built With (Copy/Paste List)

Next.js, React, Tailwind CSS, TypeScript, Google Antigravity, Google Gemini, Dexie.js, MediaPipe Face Mesh, Web Audio API, IndexedDB