Skip to content

Arjo216/Found-Match

BranchesTags

Repository files navigation

🌐 FoundMatch

The Zero-Trust, AI-Driven Institutional Capital Platform

Next.js FastAPI PyTorch PostgreSQL Encryption

FoundMatch is a professional, institutional-grade platform that connects startup founders with venture capital. It combines state-of-the-art Graph Neural Networks for matchmaking with military-grade Zero-Knowledge encryption for secure deal-flow communications.

🚀 Core Platform Architecture

FoundMatch moves beyond traditional directories by operating as a secure Institutional Operating System.

🧠 The Intelligence Layer (Hybrid AI)

  • Graph Neural Networks (LightGCN): Predicts high-probability matches based on network interaction history and collaborative filtering.
  • Semantic Analysis (Sentence-BERT): Reads pitch decks and investor theses to score contextual alignment (0-100% Global Match Index).
  • Ephemeral Executive Co-Pilot: An in-memory LLM that reads encrypted pitch decks, provides strategic critiques, and instantly destroys the file from server RAM to maintain absolute data privacy.

🛡️ The Security Perimeter (Zero-Trust)

  • Zero-Knowledge Deal Rooms: End-to-End Encrypted (E2EE) WebSockets. Messages and files are encrypted in the browser using hybrid RSA-2048 and AES-256. The backend database only stores cryptographic noise.
  • PostgreSQL Row-Level Security (RLS): A strict "Default Deny" database posture. Data APIs are completely blacked out to the public internet.
  • The Sentry Autonomous Defense: Custom ASGI middleware that actively blocks scraping, volumetric API attacks, and unauthorized payload anomalies.

📊 The Deal Flow CRM

  • Institutional Dashboards: Real-time data aggregation displaying profile views, active negotiations, and trajectory algorithms.
  • Kanban Pipeline: A native HTML5 drag-and-drop board to track relationships from Sourced to Term Sheet to Closed.

🏗️ System Architecture

flowchart TB
    %% ==========================================
    %% 🎨 CUSTOM THEME & STYLING (Dark Mode FinTech)
    %% ==========================================
    classDef user fill:#0f172a,stroke:#38bdf8,stroke-width:2px,color:#fff,rx:10px,ry:10px
    classDef frontend fill:#1e1b4b,stroke:#8b5cf6,stroke-width:2px,color:#fff,rx:8px,ry:8px
    classDef security fill:#4c0519,stroke:#f43f5e,stroke-width:2px,color:#fff,rx:8px,ry:8px
    classDef backend fill:#020617,stroke:#6366f1,stroke-width:2px,color:#fff,rx:8px,ry:8px
    classDef ml fill:#064e3b,stroke:#10b981,stroke-width:2px,color:#fff,rx:8px,ry:8px
    classDef db fill:#451a03,stroke:#f59e0b,stroke-width:2px,color:#fff,rx:8px,ry:8px
    classDef external fill:#171717,stroke:#a3a3a3,stroke-width:2px,color:#fff,stroke-dasharray: 5 5,rx:8px,ry:8px
    classDef gate fill:#000000,stroke:#eab308,stroke-width:3px,color:#fef08a,rx:12px,ry:12px

    %% ==========================================
    %% 🧑‍💼 ACTORS
    %% ==========================================
    U1((🧑‍💼 Founder)):::user
    U2((💼 Investor)):::user

    %% ==========================================
    %% 🌐 FRONTEND / CLIENT TIER (Zero-Knowledge)
    %% ==========================================
    subgraph Client_Tier ["🌐 ZERO-KNOWLEDGE CLIENT (Next.js / React)"]
        direction TB
        UI["🖥️ FoundMatch UI<br/>(Kanban Pipeline & Match Grid)"]:::frontend
        Crypto["🔐 WebCrypto Engine<br/>(RSA-OAEP & AES-GCM Key Gen)"]:::security
        Copilot["🤖 Executive Co-Pilot<br/>(Floating Document Analyzer)"]:::frontend
        Ghost["👻 Chat Ghostwriter<br/>(Contextual Reply Generator)"]:::frontend

        UI <--> Crypto
        UI <--> Copilot
        UI <--> Ghost
    end

    U1 <-->|Client-Side Encryption| UI
    U2 <-->|Client-Side Decryption| UI

    %% ==========================================
    %% 🛡️ REGULATORY GATEWAY
    %% ==========================================
    KYC{"🛡️ KYC / AML Sandbox Gate<br/>(Identity Verification & Magic Numbers)"}:::gate

    Client_Tier ==>|Authenticated JWT Requests| KYC

    %% ==========================================
    %% ⚙️ BACKEND TIER (FastAPI Microservice)
    %% ==========================================
    subgraph Backend_Tier ["⚙️ FASTAPI MICROSERVICE (Python)"]
        direction TB
        REST["🔄 Core REST API<br/>(Profiles, Projects, Filtering)"]:::backend
        WS["⚡ WebSocket Manager<br/>(Real-Time Deal Rooms)"]:::backend
        Vault["🏦 E2EE Vault Router<br/>(Blind Ciphertext & File Transit)"]:::security
        AgentRouter["🧠 AI Agent Router<br/>(Prompt Sanitization & Safety)"]:::backend
    end

    KYC ==>|Auth Pass| REST
    KYC ==>|Auth Pass| WS
    KYC ==>|Auth Pass| Vault
    KYC ==>|Auth Pass| AgentRouter

    %% ==========================================
    %% 🧬 ML ENGINE TIER (PyTorch Microservice)
    %% ==========================================
    subgraph ML_Tier ["🧬 RECOMMENDATION ENGINE (PyTorch)"]
        direction TB
        InferAPI["🎯 Inference API Gateway"]:::ml
        SBERT["📝 Sentence-BERT<br/>(Semantic NLP Pitch Parsing)"]:::ml
        LightGCN["🕸️ LightGCN<br/>(User-Item Graph Connectivity)"]:::ml
        Hyb["⚖️ Hybrid Scorer<br/>(+ Premium Priority Multipliers)"]:::ml

        InferAPI --> SBERT
        InferAPI --> LightGCN
        SBERT --> Hyb
        LightGCN --> Hyb
    end

    REST <==>|Predict Request / JSON Response| InferAPI

    %% ==========================================
    %% 🗄️ PERSISTENCE TIER (Supabase)
    %% ==========================================
    subgraph Data_Tier ["🗄️ PERSISTENCE LAYER (Supabase / PostgreSQL)"]
        direction TB
        RLS{"🔒 Row Level Security<br/>(Cryptographic Query Isolation)"}:::security
        DB[("🐘 Relational Database<br/>(Users, Projects, Interactions)")]:::db
        CipherDB[("🗃️ Ciphertext Storage<br/>(Unreadable Encrypted Blobs)")]:::db
        
        RLS --> DB
        RLS --> CipherDB
    end

    REST <==> RLS
    Vault ==>|Blind Write| RLS
    WS ==>|Asynchronous Save| RLS

    %% ==========================================
    %% 🌩️ EXTERNAL APIs
    %% ==========================================
    subgraph Ext_Tier ["🌩️ EXTERNAL CLOUD COMPUTE"]
        Llama["🦙 Groq API (Llama-3-70b)<br/>(Ultra-Low Latency Inference)"]:::external
    end

    AgentRouter <==>|Sanitized System Prompts| Llama
    Copilot -.->|Diligence Queries| AgentRouter
    Ghost -.->|Encrypted Context| AgentRouter
Loading

FoundMatch: Institutional Capital Allocation Platform

FoundMatch is an elite, AI-driven networking and deal-flow platform designed to connect high-growth startup founders with institutional investors, venture capitalists, and private equity firms.

🚀 Core Platform Features

  • PyTorch Graph Neural Network (GNN): Utilizes a custom LightGCN architecture to calculate hyper-accurate "Global Match Indexes" between founders and investors based on domain, stage, and historical interaction data.
  • Zero-Knowledge End-to-End Encryption (E2EE): All Deal Room communications and file transfers (Pitch Decks, Cap Tables) are secured client-side using a Hybrid AES-GCM + RSA-OAEP cryptographic vault. The backend server cannot read message contents or decrypt files.
  • Dual AI Engines (Groq/Llama-3): * Executive Co-Pilot: A globally available, context-aware AI widget that analyzes uploaded documents and provides strategic platform guidance.
    • Chat Ghostwriter: An in-chat M&A advisor that reads recent encrypted context and suggests highly professional, tactical responses to drive deal flow.
  • Regulatory KYC "Soft Gate": A built-in identity verification sandbox that blocks unverified participants from initiating secure Deal Rooms, simulating institutional AML compliance.
  • Interactive Deal Flow Pipeline: A drag-and-drop Kanban board for investors to seamlessly manage their pipeline from "Sourced" to "Closed".

🏗️ Monorepo Architecture

  • /Frontend: Next.js 14, React, TailwindCSS, Framer Motion, WebCrypto API.
  • /Web-Dev 2.0 (Backend): FastAPI, PostgreSQL, SQLAlchemy, WebSockets.
  • /ml_engine: PyTorch, PyTorch Geometric, Scikit-learn.

🏗️ Monorepo Structure

Found_Match/
├── frontend/             # Next.js 14 Client (UI, WebCrypto API, WebSockets)
├── backend/              # FastAPI Server (Auth, RLS Admin, Sentry, ML Router)
├── ml_engine/            # PyTorch Pipelines (LightGCN & NLP Models)
├── data/                 # Processed Datasets & Model Weights (.pth)
└── docker-compose.yml    # Full-stack Container Orchestration

⚡ Quick Start (Docker)

Launch the entire ecosystem with a single command:

# 1. Clone the repository
git clone [https://github.com/Arjo216/Found-Match-1.0.git](https://github.com/Arjo216/Found-Match-1.0.git)
cd Found-Match-1.0

# 2. Build and launch all microservices
docker-compose up --build

Frontend runs on localhost:3000 | Backend API runs on localhost:8000

2. The Frontend (Frontend/Frontend/README.md)

This focuses on your stunning glassmorphism UI, client-side encryption logic, and responsive design.

Updated frontend README to highlight the complex WebCrypto logic and stunning Glassmorphism UI components.

🎨 FoundMatch - Frontend Application

Next.js 14 • Tailwind CSS • Framer Motion • Web Crypto API

The user-facing command center for FoundMatch. Engineered for absolute privacy, real-time communication, and a frictionless, high-fidelity user experience.

✨ Key Features

  • Institutional Glassmorphism UI: A premium, dark-mode design system utilizing backdrop-blur, complex gradient meshes, and responsive CSS grids.
  • Zero-Knowledge Client (E2EE): Integrates the native browser Web Crypto API. Generates and stores RSA public/private key pairs locally to encrypt chat streams and binary file buffers before they ever touch the network.
  • Real-Time Deal Rooms: Seamless WebSocket integration for instant, encrypted founder-investor communications.
  • Interactive Kanban CRM: Native HTML5 Drag-and-Drop deal flow management (Sourced ➡️ Term Sheet ➡️ Closed).
  • Executive AI Co-Pilot: A stunning, centralized modal interface for interacting with the platform's AI, complete with context-aware smart suggestions and file staging.

FoundMatch Frontend: Next.js & Client-Side Crypto

A high-performance, institutional-grade user interface built with React, Next.js, and TailwindCSS. It acts as a "Zero-Knowledge Client," handling all data decryption and AI formatting locally.

🛡️ Security & Cryptography (/lib/crypto.ts)

  • Key Generation: Generates RSA-OAEP public/private key pairs locally in the browser upon registration.
  • Message Encryption: Uses the recipient's Public Key to encrypt message text before it ever touches the network.
  • File Encryption: Converts PDFs and documents into ArrayBuffers, encrypts them with a dynamic AES-GCM key, and encrypts that key with the recipient's RSA Public Key (Hybrid Encryption).
  • Fallback Safety: Built-in try...catch protocols to gracefully display "Legacy Unencrypted Messages" without crashing the React application.

🎨 Advanced UI/UX Components

  • KYCModal.tsx: A Glassmorphism soft-gate that intercepts users before they enter a Deal Room, forcing simulated identity verification.
  • AICoPilot.tsx: A persistent, floating AI widget with document-upload capabilities, utilizing <AnimatePresence> for smooth, state-driven transitions.
  • ChatWindow.tsx: A secure messaging interface featuring dynamic AI suggestions (The Ghostwriter), file-attachment indicators, and offline PDF dossier exporting.
  • network.tsx: An interactive, drag-and-drop Kanban board for visual Deal Flow management, integrated directly with the Deal Room chat.

🛠️ Run Locally

npm run dev
# Note: Ensure the FastAPI backend is running on port 8000, 
# and the Groq API key is valid on the server-side.

🏗️ The "Pro" Tech Stack

⚡ Getting Started (Local Development)

1. Prerequisites

Ensure you have Node.js (>= 18.x) and pnpm installed.

npm install -g pnpm

2. Installation & Execution

# Navigate to the frontend directory
cd frontend

# Install dependencies
pnpm install

# Start the development server
pnpm run dev

Navigate to http://localhost:3000 to view the application.

3. 🧹 Cache Management

If you make significant UI architectural changes, clear the Next.js cache to force a Tailwind recompilation:

rm -rf .next
pnpm run dev

3. The Backend (backend/README.md)

This focuses on your high-performance Python code, Sentry security, AI routing, and the database fortress.

⚙️ FoundMatch - Backend & AI Core

FastAPI • PyTorch • SQLAlchemy • PostgreSQL (RLS)

The high-performance, highly-secured nervous system of FoundMatch. Responsible for asymmetric data routing, zero-trust storage, and serving the PyTorch ML pipelines.

🚀 Core Infrastructure

  • Asynchronous FastAPI: Built on Starlette and Uvicorn for non-blocking, high-throughput execution (WebSockets & REST).
  • The Sentry Perimeter: A custom ASGI middleware layer that autonomously detects and throttles volumetric attacks, unauthorized scrapers, and malformed payload injections.
  • Database Fortress (Row-Level Security): Integrates with Supabase/PostgreSQL using strict RLS policies. The public API is entirely blacked out; all data is accessed securely via backend SQLAlchemy admin connections.
  • Zero-Knowledge Vault Router: The backend cannot read user messages or files. It acts as a blind courier, routing AES/RSA encrypted binary blobs (VAULT_META) between authorized UUIDs.

🧠 The Intelligence Pipeline

  • Live Analytics Engine: Dynamically calculates Global Match Indexes, active deal flows, and profile trajectory metrics via complex SQL joins.
  • Ephemeral Document AI: Utilizes PyPDF2 to read uploaded Pitch Decks entirely in system RAM, feeds the text to an LLM context window for strategic analysis, and instantly purges the file to maintain institutional privacy.
  • Hybrid AI Scoring: Hosts the endpoints that trigger the backend ml_engine (LightGCN and Sentence-BERT) for matchmaking.

FoundMatch Backend: FastAPI & AI Engine

This is the central nervous system of FoundMatch, built for high concurrency, real-time WebSocket communication, and heavy Machine Learning inference.

🧠 Recent Architectural Upgrades

  • AI Agent Router (/routers/agent.py): * Integrates the Groq API (Llama-3-70b) for ultra-low latency AI inference.
    • Powers the generate-question, chat-assist, and analyze-document endpoints.
    • Features aggressive server-side JSON cleaning and DB rollback protection against AI hallucinations.
  • KYC Sandbox Router (/routers/kyc.py):
    • Simulates third-party identity verification (e.g., Setu/Digilocker).
    • Accepts "Magic Numbers" (e.g., ABCDE1234F) to auto-verify accounts in development.
    • Masks sensitive data (e.g., XXXXX1234X) before writing to PostgreSQL.
  • Institutional Vault (/routers/vault.py):
    • Handles the secure transit of encrypted .bin files.
    • The backend stores the files but does not possess the cryptographic keys to read them.
  • Real-Time Deal Rooms (/routers/chat.py):
    • Manages active WebSocket connections for instant messaging.
    • Stores ciphertext and initialization vectors in the DB for asynchronous retrieval.

🔑 Environment Requirements

Make sure your .env includes:

DATABASE_URL=postgresql://user:password@localhost/foundmatch
SECRET_KEY=your_jwt_secret
GROQ_API_KEY=gsk_your_api_key_here
ENV=development

🛠️ Tech Stack

  • Language: Python 3.10+
  • Framework: FastAPI
  • ORM: SQLAlchemy
  • Database: PostgreSQL
  • Document Processing: PyPDF2
  • Authentication: JWT (Stateless) + Bcrypt

⚡ Getting Started (Local Development)

1. Prerequisites

  • Python 3.10+ installed
  • Access to a PostgreSQL instance (e.g., Supabase)

2. Environment Setup

# Navigate to the backend folder
cd backend

# Create the virtual environment
python -m venv venv

# Activate (Windows)
venv\Scripts\activate
# Activate (Mac/Linux)
source venv/bin/activate

3. Install & Run

# Install required Python packages
pip install -r requirements.txt

# Start the Uvicorn ASGI server
uvicorn main:app --reload --port 8000

Visit http://localhost:8000/docs to view the interactive Swagger API documentation.

About

An institutional Deal Flow OS connecting founders and investors. Built on a Next.js/FastAPI microservices architecture, it features a PyTorch LightGCN/SBERT recommendation engine, a Zero-Knowledge E2EE Vault for secure data, and Groq-powered GenAI for automated due diligence.

Topics

docker machine-learning natural-language-processing cryptography nextjs pytorch fintech recommender-system e2ee graph-neural-networks fastapi lightgcn supabase sentencebert llama-3

Resources

Readme

License

Apache-2.0 license
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages