RUMI - AI-Powered Professional Networking Platform
An AI-powered networking platform that connects people based on skills, services, and synergies rather than just titles or companies. Features semantic matching, interactive AI chat, letter writing, and event management.
The Challenge
Users spend hours manually searching for people with specific expertise or services, and there's no intelligent way to discover connections beyond keyword matching. The networking process is time-consuming, lacks context awareness, and doesn't leverage AI to understand user intent and preferences.
The Solution
Built an AI-powered networking platform that uses natural language processing and semantic embeddings to understand user profiles and preferences at a deeper level. The system extracts meaningful attributes from resumes using Google Gemini AI, creates vector embeddings for semantic search, and uses a hybrid matching algorithm that combines semantic similarity with weighted boosts for shared attributes.
Implementation Details
RUMI is a comprehensive AI-driven networking platform that revolutionizes how professionals connect. The system uses Google Gemini AI to analyze user profiles from resumes, extracting skills, industries, interests, and personality traits. It employs a hybrid matching algorithm combining semantic similarity (Sentence Transformers embeddings) with weighted boosts for shared skills, industries, and location. The platform features an interactive AI chat interface where users describe who they're looking for, and the AI asks clarifying questions before finding matches.
Technologies Used
Skills Applied
Results Achieved
Successfully built a production-ready AI networking platform with intelligent matching that goes beyond keyword search. The system can extract meaningful attributes from resumes, understand user intent through conversational AI, and provide contextually relevant matches. Real-time messaging and notifications enable seamless communication. The hybrid matching algorithm provides more accurate results than pure semantic search or pure attribute matching alone. The platform is deployed on Google Cloud Run (backend) and Netlify (frontend) with proper CI/CD pipelines.
Project Walkthrough
Challenges Faced
- ▸Integrating multiple AI services (Gemini API, Sentence Transformers) with proper error handling and fallbacks
- ▸Handling real-time WebSocket connections for chat, notifications, and presence tracking across multiple users
- ▸Processing large PDF resumes and extracting structured data reliably using AI
- ▸Managing asynchronous AI operations (letter matching, event matching) without blocking the main API
- ▸Implementing proper timezone-aware timestamps across frontend and backend
- ▸Coordinating file uploads to Google Cloud Storage with proper CORS configuration
Solutions Implemented
- Used ThreadPoolExecutor for CPU-intensive AI operations (letter/event matching) to prevent blocking the FastAPI event loop
- Built WebSocket router with connection management, presence tracking, and real-time message broadcasting to online users
- Created robust JSON extraction utilities with multiple fallback strategies for parsing AI responses with marker delimiters
- Implemented user embedding system that generates and stores vector embeddings in MongoDB for fast semantic search
- Implemented Redux store for managing chat state, matches, and WebSocket connections across the React app
- Created comprehensive error handling with try-catch blocks, logging, and graceful degradation for offline scenarios
Explore More Solutions
yield · AI Memory Assistant
Full-stack second brain that unifies streaming AI chat, persistent memory management, file-based RAG, and smart reminders. Built with FastAPI, Next.js, xAI Grok-3, and Supabase vector store.
View Case StudyRAG-Powered FAQ Support Bot
Full-stack intelligent document Q&A system using RAG (Retrieval-Augmented Generation) that enables users to query PDF and text documents with AI-powered responses, featuring automatic document indexing, vector search, and source citation.
View Case Study