EchoMind

What took a financial analyst 4 hours now takes 90 seconds.

Comment

🎙️ EchoMind

Real-Time Earnings Call & Media Intelligence Engine

💡 What took a financial analyst 4 hours now takes 90 seconds.

Upload any earnings call video and EchoMind produces a fully structured intelligence brief — who said what, extracted financial entities, speaker confidence analysis, fact-checked claims with sources, visual slide analysis, and an AI-generated executive summary.

🚀 What It Does

EchoMind transforms raw earnings call recordings into actionable intelligence by running 4 AI analysis services in parallel:

Layer What It Does
👁️ Visual Intelligence Analyzes video frames — detects slides, charts, speaker expressions, and extracts on-screen data using Reka Vision API
🎤 Voice Analysis Tracks speaker confidence, stress indicators, and emotional tone throughout the call using Modulate
🏷️ Entity Extraction Pulls structured entities — people, companies, financial figures, forward-looking statements, risk factors — using Fastino/GLiNER2 with Pioneer fine-tuning
✅ Fact-Checking Verifies claims against SEC filings, news, and public data in real-time using Yutori Research API

🎯 The Problem

Financial analysts spend 3-5 hours per earnings call manually:

  • Transcribing and reading through hour-long calls
  • Extracting key metrics and forward-looking statements
  • Cross-referencing claims against public filings
  • Summarizing findings for investment committees

EchoMind automates the entire workflow.

🖥️ Demo

Input: Paste a YouTube URL of any public earnings call

Output:

  • 📊 Executive Summary — 3-paragraph AI brief with key takeaways
  • 🏷️ Entity Panel — Companies, people, financial figures, risk factors extracted and categorized
  • 📈 Voice Confidence Timeline — Interactive chart showing speaker confidence over time (spot when the CEO gets nervous during the margin question!)
  • 👁️ Visual Insights — Slides and charts detected with content extracted ("Revenue slide at 14:23 shows 23% YoY growth")
  • ✅ Fact Checks — Each major claim tagged as Verified, Needs Context, Disputed, or Unverified with sources

🏗️ Architecture

┌──────────────────────────────────────────────────────────────┐
│                     Render Platform                           │
│                                                               │
│  ┌──────────────┐  ┌───────────────┐                        │
│  │ Static Site   │  │  Web Service  │                        │
│  │ React + Vite  │──│  FastAPI      │                        │
│  │ Dashboard     │  │  Backend API  │                        │
│  └──────────────┘  └───────┬───────┘                        │
│                            │                                  │
│                     ┌──────┴───────┐                         │
│                     │  PostgreSQL   │                         │
│                     │  Results DB   │                         │
│                     └──────────────┘                         │
└──────────────────────────────────────────────────────────────┘
                             │
               ┌─────────────┼─────────────┐
               │             │             │
      ┌────────▼───┐  ┌─────▼──────┐  ┌──▼──────────┐
      │ Reka       │  │ Fastino    │  │ Yutori      │
      │ Vision API │  │ GLiNER2    │  │ Research    │
      └────────────┘  └────────────┘  └─────────────┘
               │
      ┌────────▼───┐
      │ Modulate   │
      │ Voice API  │
      └────────────┘

3 Render services: Web Service + Static Site + PostgreSQL

⚡ Quick Start

Backend

cp .env.example .env
# Add your API keys to .env

cd backend
uv venv
uv pip install -r requirements.txt
uv run python main.py

Frontend

cd frontend
pnpm install
pnpm dev

🚢 Deploy to Render

Connect this repo and Render auto-deploys all 3 services using render.yaml — infrastructure as code.

🔧 Tech Stack

  • Backend: Python 3.11, FastAPI, SQLAlchemy, PostgreSQL
  • Frontend: React 18, Vite, Tailwind CSS, Recharts
  • AI Services: Reka Vision, Modulate, Fastino/GLiNER2, Yutori
  • Infrastructure: Render (Web Service, Static Site, PostgreSQL)
  • Media Processing: yt-dlp, ffmpeg

📂 Project Structure

├── render.yaml              # Render infrastructure-as-code (3 services)
├── backend/
│   ├── main.py              # FastAPI application
│   ├── services/
│   │   ├── orchestrator.py  # Parallel analysis pipeline
│   │   ├── reka_service.py  # Visual intelligence
│   │   ├── modulate_service.py  # Voice analysis
│   │   ├── fastino_service.py   # Entity extraction
│   │   └── yutori_service.py    # Fact-checking
│   ├── models/              # SQLAlchemy + Pydantic schemas
│   └── routers/             # API endpoints
└── frontend/
    └── src/components/      # React dashboard components

🔑 API Integrations

Reka Vision API 👁️

Multimodal video/image understanding — analyzes video frames for slides, charts, speaker expressions. Enterprise-grade visual AI for financial content analysis.

Modulate 🎤

Voice intelligence — analyzes speaker tone, prosody, and confidence. Goes beyond speech-to-text to understand how things are said. Detects when executives sound evasive during tough Q&A.

Fastino/GLiNER2 🏷️

Fast, structured entity extraction with a 205M-parameter model. Extracts financial entities (companies, metrics, forward-looking statements, risk factors) and classifies statement types. Fine-tuned on financial data using Pioneer for improved F1 scores.

Yutori 🔍

Autonomous web research agents — verifies financial claims against SEC filings, news, and public data.

Render ☁️

Modern PaaS powering the full stack — 3 service types deployed via infrastructure-as-code (render.yaml). Web service, static site, and managed PostgreSQL.

Built With

  • modulate
  • reka
  • render
  • yuktori
Share this project:

Updates