bridGe

Framing the Problem

Modern social media platforms prioritize engagement over genuine connection, making it difficult for people to reconnect with lost friends, colleagues, or acquaintances in a meaningful way. At a time when 1 in 4 adults feel socially disconnected, we need better tech to help people rebuild and sustain meaningful relationships.

Key Issues:

1. Limited search capabilities – Platforms like Facebook & LinkedIn only let you search by name or location, making it hard to find someone if you don’t have exact details.

2. No AI-powered people-finding tools – Google Reverse Image Search doesn’t connect to social profiles, and paid people-finder tools are ineffective, especially globally.

3. Lack of genuine and natural reconnection features – Even if you do find someone, there’s no built-in way to rekindle and maintainthe relationship beyond likes and superficial interactions.

Idea Explanation

Our platform helps people find, reconnect, and maintain relationships with lost friends, colleagues, and acquaintances in a meaningful way. Unlike traditional social networks, we prioritize intentional connection over engagement metrics like likes and shares.

How It Works:

1. Smarter Search – Users can find people using a combination of:

   • AI-powered facial recognition (voluntary opt-in) powered by user's photo submission

   • Name, location, school, job history, and shared interests

   • Smart filtering to refine results and save time searching endlessly

2. Seamless Reconnection – Once you find someone:

   • Send a reconnection request

   • If accepted, start private conversations (not public feeds)

   • Share updates, messages, and personal milestones

3. AI-Powered Relationship Maintenance – Stay in touch with:

   • AI-generated check-ins & conversation starters

   • Effortless scheduling with calendar sync & meetup suggestions

   • Personalized notifications that encourage meaningful interactions

4. Gamification for Consistency – Keep relationships alive with:

   • Weekly streaks for ongoing conversations

   • Rewards & discounts for attending events together

   • Fun challenges that encourage real-world reconnections

Why It’s Different:

AI + Smart Search – Goes beyond basic name/location searches with DeepFace AI for facial recognition + multi-criteria filtering.
Privacy-First – Users have full control over visibility, opt-in features, and the ability to delete their profiles anytime.
*Designed for Meaningful Reconnection – Focuses on real relationships, not endless scrolling or vanity metrics.

Implementation

Used wireframe to show the User Interface

To build our AI-powered reconnection platform, we propose using a Flask backend, Vanilla JavaScript, HTML, and CSS for the frontend, and AWS DynamoDB as our database. This setup ensures simplicity, scalability, and ease of maintenance while integrating powerful AI-driven search and reconnection features.

1. Tech Stack

Frontend (User Interface)

• HTML, CSS, JavaScript (Vanilla JS) – A lightweight approach without the overhead of frameworks

• AJAX / Fetch API – Handles communication with our Flask backend

• CSS for Styling – Ensures flexibility in UI design

Backend (Business Logic & APIs)

• Flask (Python) – A lightweight, flexible framework for handling API requests and user interactions

• Flask-RESTful – Organizes our API endpoints efficiently

• Flask-SocketIO – Supports real-time messaging and notifications

• OpenAI’s GPT-4 API – Generates AI-powered conversation starters and engagement reminders

Database & Storage

• AWS DynamoDB – A NoSQL, scalable solution for managing user profiles, connections, and interactions

• DynamoDB Global Secondary Indexes (GSIs) – Enables fast, multi-criteria search queries

• AWS S3 – Stores profile pictures and facial recognition data

• Redis (or DynamoDB storage) – Caches AI-generated facial embeddings for fast lookup

2. Key APIs & Services

AI-Powered Search (Facial Recognition & Smart Filtering)

• DeepFace (Python-based AI model) – Extracts facial embeddings and enables image-based searches

• DynamoDB GSIs – Allows users to search by name, location, job history, and other structured criteria

• Custom Search API (Flask) – Combines AI-powered facial matching with database filtering

Reconnection & Communication

• Flask-SocketIO – Handles real-time messages and notifications

• Twilio API – Provides optional SMS and email notifications for reconnection reminders

• Google Calendar API – Syncs with users' schedules to facilitate meetups

• GPT-4 API (OpenAI) – Suggests conversation starters and personalized engagement reminders

Privacy & Security

• OAuth (Google, Facebook, LinkedIn) via Flask-OAuthlib – Ensures secure user authentication

• End-to-End Encryption (AES-256 or Flask-Talisman) – Protects private messages and user data

• GDPR Compliance – Gives users control over visibility, opt-in features, and data deletion

3. Architecture Flow

• User Uploads Information & Images – The frontend collects profile data such as name, location, interests, and an optional profile image.

• Flask Backend Processes Search Requests –

  • If an image is uploaded, DeepFace extracts facial embeddings and searches against stored data.

  • If it's a text-based search, DynamoDB GSIs filter results based on multiple criteria.

• User Finds Connection & Sends Request – Once a match is found, the user can send a reconnection request.

• If Accepted, Additional Features Unlock – Messaging, AI-powered check-ins, and meetup scheduling become available.

• Gamification & Engagement Features – Streaks, AI-generated check-ins, and event-based rewards encourage ongoing reconnection.

4. Deployment & Scaling

• AWS EC2 or AWS Lambda – Runs the Flask backend efficiently

• AWS S3 – Handles static asset and image storage

• GitHub Actions + Docker – Automates our CI/CD pipeline

• AWS CloudWatch + Prometheus – Monitors performance and logs key events

Challenges

What did you struggle with? How did you overcome it?

Connecting

• Access to APIs for gloabl search: We struggled to get access to Linkedin and Facebook's APIs for our global search algorithm, so we went with the Github API to find user information outside of our database

• Access to free-tiers for Mongodb: For our database, we used Mongodb's free tier, which prevented us from creating user profiles with images (GridFS conversions) to test our image search logic

• Time restrictions prevented us from building a fully-functional front-end and connecting it to the backend logic that works at the moment

Accomplishments

What did you learn? What did you accomplish?

• We learnt how to work together in a team with completely new people.

• We managed to figure out the most part of our logic and implement part of it

• We did comprehensive research to see the legal implications of the kind of data we need, and the limitations of existing public APIs in that regard

• We came up with a user interface for our MVP and settled on design for future work

• Different technologies

Next Steps

What are the next steps for your project? How can you improve it?

• Completing the implementation of the flask backend and frontend, and integrate them

• Complete testing

• Investigate current AI tools that exist that can be incorporated.

Essential Links

Presentation:

https://www.canva.com/design/DAGhJ5PkYkM/GWaHfcuLmS57OoE6Tq_vLA/view?utm_content=DAGhJ5PkYkM&utm_campaign=designshare&utm_medium=link2&utm_source=uniquelinks&utlId=h8d64e7afdf

Wireframe:

https://www.canva.com/design/DAGhJQ8ft9k/wti42PPfRbv2Wt1v9J4oGQ/view?mode=prototype