RapportAI

• 

  Homepage of RapportAI

Inspiration

RapportAI started with a simple but powerful insight during the Bolt AI Hackathon: people struggle to connect—not because they lack empathy or intelligence, but because they’re unsure how to begin. Whether it’s a dating app, a networking DM, or a sales pitch, most people feel anxiety about saying the wrong thing, especially across cultural or language lines.

Our team—made up of tech interns passionate about AI, psychology, and UX—realized we all had one thing in common: we’ve hesitated to send that first message. That hesitation costs relationships, opportunities, and confidence. So we built RapportAI: an AI-powered relationship advisor that helps people initiate meaningful conversations with less stress and more success.

What it does

RapportAI is like a smart wingman that analyzes public social media profiles (starting with Instagram, expanding soon to LinkedIn and Facebook), infers personality, interests, and tone, and then generates personalized message suggestions—whether it’s a clever icebreaker, a smooth reply, or a professional intro.

It's designed to work across contexts—social, dating, networking, even sales—and supports both English and Thai, with culturally adapted phrasing. It’s fast, human-like, and scalable.

How we built it

• Frontend: Developed with Bolt and deployed on Netlify for responsiveness and speed.
• Backend: Built with FastAPI and deployed to Hugging Face Spaces, it handles profile analysis, request routing, and API integration.
• Database: Supabase stores user sessions, feedback, and personalization settings.
• APIs:
  • Google Gemini API: Powers AI-generated message responses with natural tone control.
  • APIFY API: Extracts public Instagram data (bios, captions, hashtags) for contextual understanding.

We structured the app with scalability in mind, allowing future support for Twitter, LinkedIn, and custom CRM integrations.

Challenges we ran into

• Ethical data usage: We had to carefully design scraping logic to respect user privacy and platform terms.
• Language switching: Training AI to write in both Thai and English while respecting tone, formality, and nuance was tough.
• UX simplicity: We iterated heavily on how to show suggestions without overwhelming the user.

Accomplishments that we're proud of

• Created a working MVP that goes from social profile → personality extraction → chat suggestion in one smooth flow.
• Achieved natural-feeling message generation in both Thai and English.
• Clear use cases emerged for both consumers and professionals (salespeople, recruiters, freelancers).

What we learned

• Personalized communication isn't a luxury—it's an expectation. People engage better when messages feel relevant and thoughtful.
• Empathy and tone are the hardest things to teach a machine—but not impossible.
• Localization is more than translation: building for cultural context is critical.
• A tool that empowers people to express themselves better has broad emotional and commercial impact.

What's next for Untitled

• B2B API development: Enabling CRM, HR, and sales platforms to fetch personality insights directly.
• Expansion: Adding support for more platforms (LinkedIn, Twitter) and more languages.

Our mission is simple: help the world communicate better—one personalized message at a time.

Built With

• apify
• bolt
• fastapi
• gemini
• huggingface
• netlify
• scraping
• supabase