VisaVerse | AI Mobility Copilot

Landing Page
Traveller Details Form
Context Form
Copilot Strategy Page1
Copilot Strategy Page2
AI Document Analyzer

Inspiration

Moving across borders is broken. Whether you are a digital nomad, a student, or an HR manager handling relocation, the process is plagued by confusing government websites, expensive legal consultations, and the constant fear of rejection due to a single missing document.

We asked ourselves: "What if you had a VISA Officer in your pocket?" Not just a chatbot, but a strategic Copilot that simulates approval odds, identifies risks before they happen, and builds a project management timeline for your move. That’s how VisaVerse was born.

What it does

VisaVerse is an AI-powered global mobility intelligence engine.

Mobility Score™: Instantly calculates your eligibility score (0-100) based on age, income, role, and destination.

Approval Simulator: Uses AI to simulate a visa officer's decision-making process, highlighting concrete rejection risks.

Smart Timeline: Generates a realistic, step-by-step Gantt-chart style roadmap from "Application" to "Landing".

AI Document Screener: Users can paste their CV or Offer Letter text, and the AI scans it for red flags (e.g., "Salary too low for Blue Card").

Interactive 3D Visualization: A stunning, interactive globe visualization to visualize migration corridors.

How we built it

We wanted the UI to feel like a "Sci-Fi Cockpit" for global mobility.

Frontend: Built with Next.js 16 (Turbopack) for speed. We used Aceternity UI and Shadcn/UI components to achieve the "Glassmorphism" and "Cyberpunk" aesthetic. The 3D elements are powered by react-globe.gl and framer-motion for buttery smooth transitions.

Backend: A robust Node.js/Express server handles rate limiting and API orchestration.

AI Engine: We utilized OpenRouter (LLM) with extensive Prompt Engineering. We created a strict JSON-enforced system prompt that forces the AI to act like a conservative immigration lawyer, ensuring the output is structured and parsable for our UI dashboard.

Challenges we ran into

Taming the AI: LLMs love to chat. Making them output strict, complex JSON schemas for our "Bento Grid" dashboard without breaking the frontend was tough. We had to write a robust safeExtractJSON parser to handle edge cases where the AI added markdown or conversational filler.

The Globe Resize Glitch: Integrating a WebGL globe inside a responsive, scrollable layout caused massive layout shifts and re-rendering loops. We solved this using React.memo and careful CSS containment to ensure high performance even on lower-end devices.

Next.js 16 Compatibility: Being on the bleeding edge meant dealing with Turbopack caching issues, which we resolved by optimizing our build configuration.

Accomplishments that we're proud of

Building a dashboard that actually looks like a SaaS product, not just a hackathon prototype.

The Document Checker feature works incredibly well—it actually catches subtle errors like "Job title doesn't match education".

Achieving a 98% success rate in parsing complex JSON responses from the AI.

What we learned

Prompt Engineering is Engineering: We learned that getting reliable data from LLMs requires treating English like code. Strict schema enforcement and fallback parsing logic were crucial to prevent "JSON parse error" nightmares.

Performance is a feature: Integrating high-fidelity 3D elements (The Globe) in a React app taught us valuable lessons about rendering cycles. We mastered React.memo to stop components from re-rendering unnecessarily.

Bleeding Edge has sharp edges: Using Next.js 16 with Turbopack was risky. We learned to navigate beta-software documentation and debug cryptic build errors, which made us better developers under pressure.

The "Happy Path" doesn't exist: Real-world inputs are messy. Building the Document Screener taught us to handle edge cases where user input is vague or unstructured, making our system more resilient.