Inspiration
My inspiration for UniPilot came directly from my own experience. As an international student, I spent years dreaming of studying abroad. I handled the forms, the transcripts, and the applications. But the most terrifying barrier was the F-1 Visa Interview. The interview is short, high-stakes, and often conducted by an officer with a poker face. There were no realistic tools—just static FAQs that couldn't prepare me for a dynamic, probing conversation.
What it does
Real-Time Interview Simulation: The core feature uses an ElevenLabs Voice Agent to simulate a formal, adaptive U.S. Visa Officer. It listens to your spoken answers and asks probing follow-up questions based on actual visa law.
AI Critique and Score: After the interview, the Agent provides a Visa Confidence Score and a detailed, personalized critique (e.g., "Weak on financial ties, strong career path"). This is objective, actionable feedback.
Animated Journey Tracking: The Roadmap screen uses a live animation of a plane flying from the user's home country to their destination. The plane's progress is directly linked to their checklist completion percentage, gamifying the journey.
End-to-End Management: The app manages all aspects of the journey, including a modular Checklist, a Documents hub for uploads, and a Memories tab for celebrating milestones.
How I built it
Tech Stack: React Native (Expo), Bolt Native Database, ElevenLabs Agents SDK.
Core AI Integration (ElevenLabs): I created a specialized ElevenLabs Agent persona (the Visa Officer) and provided it with a domain-specific knowledge base on F-1 visa law (e.g., the 'Ties to Home' doctrine). I integrated the Agent using the ElevenLabs SDK to manage the low-latency, real-time voice connection.
Data Persistence: I utilized the easy-to-implement Bolt Native Database for user authentication, storing user profile data, all checklist progress, and saving every Interview Session (score and critique).
Design: I enforced a global Pink and White Minimalist design system across all five tabs to create a visually inviting and stress-reducing user experience.
Challenges I ran into
Animation Logic: Implementing the unique Animated Progress Map was complex. It required carefully structuring code to link a database variable (the progress percentage) to a visual property (the plane's translateX position) using the native React Native Animated API.
Context Management: While the conversation was smooth, ensuring the LLM agent maintained the precise 5-question sequence and then switched personas for the final critique required constant refinement of the system prompts to keep the conversation on track.
Accomplishments that I'm proud of
The Working Voice Simulation: The fact that the core, high-fidelity, real-time interview works flawlessly on the mobile emulator is my proudest achievement. This is a game-changing feature that demonstrates a winning use of the ElevenLabs Agents Platform.
Visual Cohesion: Achieving the ambitious Pink and White Minimalist Design and the fluid Animated Map showcases a high level of polish and attention to User Experience in a short time.
What I learned
The most valuable lesson was that prompt engineering is now part of the application architecture. I learned that the quality of the AI feature depends less on code written manually and more on the precise, structured instructions given to the LLM. Switching to Bolt's Native Database also taught me the immense value of prioritizing easy-to-integrate tools to maximize feature delivery in a tight time window.
What's next for UniPilot - Study Abroad AI App
Maybe a few more days of vibe-coding and I see UniPilot becoming a very popular app amongst International students who are considering to apply for U.S. universities. Millions of international students apply for F1 Visa each year. This is a very big market to explore.
Built With
- bolt
- database
- elevenlabs
- expo.io
- react-native
Log in or sign up for Devpost to join the conversation.