DEV Community: Akhil Mani

Universal Health Guardian : lost GitHub account but rebuild,with Chatgpt

Akhil Mani — Tue, 26 May 2026 00:10:02 +0000

Universal Health Guardian
–

Rebuilding from Scratch After Losing Everything

This is a submission for the GitHub Finish-Up-A-Thon Challenge

What I Built

I built Universal Health Guardian (UHG) — an AI-powered healthcare assistant platform focused on helping patients and healthcare providers monitor health risks, visualize patient data, and generate predictive insights using machine learning.

The project combines:

A modern React dashboard UI

A Node.js + Express backend

PostgreSQL database architecture

ML prediction endpoints

Real-time health analytics concepts

AI-assisted workflows powered through GitHub Copilot and vibe coding sessions

The idea started as a simple healthcare dashboard experiment, but evolved into a much larger vision: a universal intelligent healthcare companion that could eventually support hospitals, remote clinics, and individual patients.

Some of the core systems we built include:

Patient health tracking dashboards

Risk prediction APIs

Medical data visualization

AI-assisted recommendation flows

Modular monorepo architecture for scalability

Frontend-backend ML integration structure

This project became more than just code for me.

A major setback happened when my GitHub account got suspended unexpectedly. I lost access to years of work, repositories, experiments, and unfinished builds. I tried contacting support multiple times, but my emails never properly connected through the system.

For a while, it genuinely felt like everything was gone.

Instead of stopping, I started rebuilding from memory using AI-assisted development workflows and late-night vibe coding sessions. Universal Health Guardian became the comeback project that helped me regain momentum.

Demo

Features

Interactive healthcare dashboard

AI-assisted patient analytics

Prediction endpoints architecture

Scalable monorepo setup

Responsive frontend with Tailwind CSS

REST APIs for patient data and ML integration

Tech Stack

React 18

Tailwind CSS

Node.js

Express.js

PostgreSQL

Machine Learning integration concepts

GitHub Copilot

AI-assisted development workflows

Walkthrough

Added screenshots:

Dashboard homepage
Patient analytics panel
API response preview
System architecture diagram
Mobile responsive view

The Comeback Story

Before this challenge, the project existed mostly across lost repositories, incomplete builds, local experiments, and AI-generated architecture drafts.

When my GitHub account got suspended, I thought I had lost:

My repositories

Project history

Previous prototypes

Collaboration trails

Deployment experiments

Documentation

That situation forced me to rethink how I build.

Instead of trying to perfectly recover everything, I focused on rebuilding better:

Cleaner architecture

Better scalability

Modular code structure

Faster iteration cycles

AI-assisted development

Rapid prototyping workflows

During this rebuild process, I used GitHub Copilot heavily alongside iterative vibe coding sessions to reconstruct systems, generate boilerplate faster, and experiment with architecture ideas rapidly.

What changed during the comeback:

Rebuilt frontend architecture from scratch

Reorganized backend APIs

Created structured ML integration endpoints

Improved dashboard responsiveness

Simplified deployment planning

Converted scattered ideas into a more production-oriented architecture

This challenge became less about “finishing an old project” and more about proving to myself that losing everything doesn’t mean starting over with nothing.

My Experience with GitHub Copilot

GitHub Copilot became one of the biggest reasons this rebuild moved as fast as it did.

Instead of spending hours rewriting repetitive boilerplate, I used Copilot to:

Generate API routes

Scaffold React components

Speed up backend logic creation

Refactor repetitive code

Prototype architecture ideas quickly

Explore integration patterns faster

The most valuable part wasn’t just code completion — it was momentum.

After losing access to previous repositories, motivation was honestly difficult at times. Copilot helped reduce the friction between ideas and implementation. That made it easier to keep building consistently instead of getting stuck rebuilding basic foundations.

Some areas where Copilot helped most:

Express route generation

React component scaffolding

Tailwind layout structures

API integration patterns

Backend folder organization

Rapid experimentation

This project showed me how AI-assisted development can genuinely help solo builders recover faster after setbacks.

Final Thoughts

Universal Health Guardian is still evolving, but this Finish-Up-A-Thon pushed me to turn recovery into progress.

Even after losing repositories and access to older work, rebuilding taught me:

consistency matters more than perfection,

momentum matters more than setbacks,

and sometimes the comeback version becomes stronger than the original.

If you’re rebuilding something you lost too — keep going.

The Android Spam shield

Akhil Mani — Wed, 20 May 2026 11:22:51 +0000

Trust First, Customers Later: Building the Android "Spam Shield" with Antigravity 2.0 and Android 17

A Personal Journey from Being Spammed to Building a Shield

The moment you sign up for a service, buy a new device, or share your number, it begins. The spam. I lived through this firsthand when a toxic client weaponized my information. It didn’t just feel like an annoyance; it felt like a violation of my digital sovereignty."
For too long, big tech's motto has been "Growth first, privacy later." It's time to flip the script. Our new architectural philosophy is simple: Trust first, customer later.
Watching the Google I/O 2026 keynotes, I realized that we finally have the tech stack required to move from passive call-blocking to active, autonomous defense. This is the story of how we can combine Android 17, Antigravity 2.0, and Gemini 3.5 Flash to build an aggressive, built-in Spam Shield that doesn't just block spammers—it hunts them, references a global database, and hands their data straight to law enforcement.

The 3-Tier "Spam Shield" Architecture
To completely lock down a Pixel or Android device, we have to synthesize three massive announcements from I/O 2026 into a single cohesive system:

[Android 17 OS Layer] -> [Antigravity 2.0 Global DB Check] -> [Gemini 3.5 Flash Agent] -> [Automated Legal Report]

The Gatekeeper: Android 17 OS Verification Instead of waiting for a spammer to text or call, Android 17's new deep OS verification and integrity APIs ensure that unverified apps, emulators, or bot-farm clones cannot even execute background tasks on the device network. If an incoming communication doesn't pass cryptographic verification, it's flagged instantly before it can scrape user data. 2. The Brain: Antigravity 2.0 & Gemini 3.5 Flash Sub-Agents When a suspicious interaction bypasses basic filters, an isolated Antigravity 2.0 agent is spun up instantly. Leveraging the mind-boggling speed of Gemini 3.5 Flash (clocking in at 289 output tokens/sec), the agent cross-references the spammer's metadata against Google's global security database in milliseconds. Because Antigravity 2.0 allows agents to spawn independent, asynchronous sub-agents on the fly, one sub-agent can seamlessly isolate the spammer in a virtual sandbox while another handles data compilation. 3. The Hammer: Automated Law Enforcement Reporting If the global database or the Gemini model flags the entity as a malicious spammer/scammer with a probability score greater than 95%, the shield triggers an automated legal reporting pipeline. Using secure JSON hooks, it compiles the spammer's IP header, carrier routing info, and payload, generating a standardized report sent directly to local cyber security teams or federal law enforcement (like the FBI's IC3 portal).

The Code Blueprint: The Asynchronous Spam Handler
Here is a conceptual look at how an Antigravity 2.0 agent script handles a flagged interaction using the new /goal directive to run an enforcement pipeline to completion:

// Antigravity 2.0 Autonomous Spam Shield Hook
import { googleSecurityDB, lawEnforcementAPI } from '@google/android17-security';
import { Gemini35Flash } from '@google/vertex-ai';

export async function onSuspiciousInteractionTrigger(interaction) {
// Step 1: Query Google's Real-time Database
const isFlagged = await googleSecurityDB.checkEntity(interaction.senderMetadata);

if (isFlagged || interaction.trustScore < 0.05) {
console.log("Spammer detected. Initiating 'Trust First' protocol.");

// Step 2: Spin up a Gemini 3.5 Flash Agent to analyze intent and compile evidence
const analysis = await Gemini35Flash.analyzePayload({
  content: interaction.payload,
  systemInstruction: "Extract routing, headers, and malicious intent indicators."
});

// Step 3: Trigger automated law enforcement pipeline asynchronously
await lawEnforcementAPI.submitCyberCrimeReport({
  evidence: analysis.extractedData,
  routingDetails: interaction.networkHeaders,
  classification: "Automated Mass Spam / Data Harvesting"
});

// Step 4: Vaporize the data trail on the user's local device
await interaction.blockAndPurge();

}
}

Why This Matters for the Future of Humanity
When we protect a Google user, we shouldn't just minimize a notification; we should actively dismantle the spammer's infrastructure. By forcing their data back into law enforcement databases the second they hit a honeypot, we make spamming too legally and financially risky to sustain.
Software development shouldn't just be about building cooler apps; it should be about building an ecosystem where everyday users feel entirely safe. Let's build a web where trust isn't a premium feature—it's the baseline.
Disclaimer: This post is a submission for the Google I/O 2026 Writing Challenge. I used Gemini to help synthesize my personal experience and architectural ideas into this final written piece.

Hope google users would be saved from the project we are building...
Soon in playstore free for all..

https://youtube.com/shorts/s4TdWLdX4rc?si=_4OFty8-Qvq9DuoD