mEnContre

Your next big idea starts with the right partner.

Comment

Inspiration

The spark of inspiration for the birth of mEnContre arose from a painfully common frustration experienced firsthand: the exhausting and ineffective attempt to form development teams.

Some time ago, we decided to participate in a global hackathon on the DevPost platform. With a great idea in mind, the necessary step was to find partners with complementary skills. The strategy was traditional: search for recommended profiles on DevPost, follow them, and send dozens of personalized messages on LinkedIn, email invitations, or through available contact channels. The result, however, was discouraging: the response rate was extremely low. Silence prevailed. And when they finally responded, the technical misalignment was evident—they didn't possess the exact skills required for our proposal, and we didn't meet the requirements they were looking for in their scope.

This barrier exposed a critical flaw in the ecosystem: while extraordinary ideas and talent abound, direct communication bridges and granular semantic compatibility filters do not exist in static profile directories.

mEnContre was created precisely to transform this scenario. Instead of forcing developers to blindly navigate generic feeds and saturated networks, we designed an intelligent AI that understands recruitment intentions in natural language, scans the web in real time using modern scraping engines (generating the AI Score of mutual affinity), validates precise technical suitability, and extracts direct communication channels so that the invitation to the ideal candidate happens proactively, in a qualified and friendly way, with a single click.

What it does

mEnContre is an intelligent matchmaking and engineering lead orchestration engine that operates on three main fronts:

  1. Smart Matching Based on Intent: The user describes their project in natural language (e.g., "Looking for an engineer from IIT Patna with NLP and RAG knowledge for the DeveloperWeek New Yourk Hackathon"). The system uses the Gemini 3.5 model to interpret the technical intent and compute the AI Score of compatibility from 0 to 100%, detailing the reasons for the affinity.

  2. Integration of Active Methods (Internal vs. Web): Merges the search in talent databases registered on the platform with profiles extracted in real time from external development platforms and public directories such as GitHub, Bebee, and LinkedIn.

  3. Firecrawl Lab & Contact Reverse Engineering: Provides an advanced dashboard where the user can scrape entire portfolio pages. AI analyzes the scraped page layout, structurally recognizes each person, and enables Deep Crawling. This crawling automatically scrapes the subpages of these professionals on BeBees or GitHub to find the direct link to send a chat message, email, or WhatsApp, dynamically linking this link to the "Show Interest" button in the main ecosystem.

How we built it

We built mEnContre using a modern, fully reactive and secure full-stack architecture:

  • Frontend: Implemented in React 19 with TypeScript and packaged by Vite. The design is governed by Tailwind CSS, prioritizing contrast, technical readability, and immersive dark interfaces. We use Motion/React to transparentize background AI processes through animation flows and adaptive layouts.

  • Autonomous Backend: We developed a server in Express using Node.js. It acts as a secure proxy server that manages the database in db.json, performs pagination, and isolates confidential encryption keys and API tokens from the browser.

  • Cognitive Engines (AI): We integrated the official @google/genai SDK using the Gemini 3.5-flash model to compute semantic similarity, parse profiles from structured pages, and extract dynamic links from subpages.

  • Web Crawler: We adopted the powerful data conversion API Firecrawl to extract complete pages, processing requests in optimized Markdown format so that AI can quickly map contact data with high accuracy. ## Challenges we ran into

Throughout development, we faced four major technical challenges:

  • Request Saturation and Gemini Limitations (503): Cloud call congestion sometimes caused temporary failures when interpreting very long profiles. We solved this by implementing the ForgeMatch Local Scoring Engine, an intelligent local fallback algorithm that takes over in emergencies without crashing the user interface if the Gemini API fails.

  • Identification of Hidden Contact Methods: Talent platforms like BeBee use complex and dynamic layouts to hide internal chat links or communication networks behind customer interactions. We overcame this by developing the **

Built With

Share this project:

Updates