QuantBrief
A multi-agent AI system that acts as a personal Chief Investment Officer — monitoring markets, SEC filings, and news to deliver synthesized audio briefings on schedule. Built on a 5-stage pipeline using 4 specialized Mistral models and ElevenLabs voice synthesis, the platform democratizes institutional-grade market intelligence for retail investors at a fraction of Bloomberg Terminal costs.
5
AI Models
25+
API Endpoints
5
Languages
33
Automated Tests
Key Differentiators
- 5-stage multi-agent pipeline with 4 specialized Mistral models orchestrated end-to-end
- 256K context SEC filing analysis — full 10-K in a single pass, no chunking or RAG needed
- Scheduled automated briefings with multilingual audio via ElevenLabs in 5 languages
- Production-grade infrastructure with JWT auth, PostgreSQL, rate limiting, and 33 automated tests
The Problem
150+ million retail investors face massive information asymmetry. A Bloomberg Terminal costs $25,200/year. Without it, investors are buried under 300+ financial articles per hour they cannot read, 80-120 page SEC filings in legal jargon they cannot parse, and English-only sources that exclude European investors. No tool connects the dots across these fragmented data sources into actionable intelligence.
Multi-Agent Pipeline
QuantBrief orchestrates 5 specialized AI agents in sequence: Ministral 3B screens news at ~50ms per article for high-throughput filtering, Mistral Large 3 analyzes full SEC filings in its 256K context window without chunking, Magistral Medium performs chain-of-thought reasoning about portfolio impact with confidence scores, Mistral Large 3 synthesizes cross-source intelligence into executive briefs, and ElevenLabs generates professional audio briefings in 5 languages.
Scheduled Automation
Unlike on-demand tools, QuantBrief runs fully automated brief generation on configurable schedules — every 4 hours, daily, or weekly. Users set their preferred time and ticker sources (custom tickers, portfolio holdings, watchlist, or all combined). The system uses APScheduler with PostgreSQL-backed persistence, ensuring schedules survive server restarts. A concurrency-safe asyncio lock prevents overlapping pipeline runs. Your morning brief is ready before you wake up.
Real-Time Market Data
The platform integrates live market data from multiple free sources: yfinance for quotes, candles, and financials; SEC EDGAR for regulatory filings; Finnhub for ticker search and company profiles; FRED for macroeconomic indicators; and RSS feeds for real-time news. Interactive dashboards display candlestick charts, technical analysis (RSI, MACD, Bollinger Bands), financial ratios with radar chart comparison, and per-ticker news with AI-scored sentiment.
Production Infrastructure
QuantBrief is deployed with production-grade engineering: React 19 frontend on Vercel with Zustand state management and lazy-loaded routes, FastAPI backend on Railway with async PostgreSQL and Redis caching, JWT authentication with access/refresh tokens and bcrypt hashing, per-endpoint rate limiting via SlowAPI, security headers (CSP, X-Frame-Options, Referrer-Policy), WebSocket for real-time pipeline progress, and a CI/CD pipeline running 33 automated tests on every push.
Why It Matters
QuantBrief turns the $25,200/year Bloomberg Terminal value proposition into an accessible, AI-native experience. By leveraging Mistral's 256K context window for full SEC filing analysis — eliminating the need for RAG or chunking — and orchestrating specialized models for each stage of the intelligence pipeline, the platform delivers institutional-quality market briefs that retail investors can consume as audio while commuting. Built for the Mistral AI Worldwide Hackathon 2026, the system demonstrates that multi-agent AI architectures can democratize financial intelligence.
Interested in learning more about this venture?
Get in Touch