Velarith

Enterprise financial analytics powered by Claude AI. Fuses real-time data with technical indicators, sentiment analysis & predictive insights—delivering institutional-grade intelligence for traders.

Comment

Inspiration

The inspiration for Velarith came from a simple but frustrating truth that high-quality financial analytics are locked behind expensive terminals and institutional subscriptions. As retail traders with the ability to change, we wanted to level the playing field. We were inspired by Claude Sonnet 4.5’s reasoning capabilities and realized we could merge AI interpretation with real-time market data to deliver Wall Street-grade analysis for everyone. The name Velarith represents speed and clarity. Our goal was to make complex financial data not just accessible, but understandable and actionable.

What it does

Velarith is an AI-powered financial analytics platform that fuses real-time stock data with Claude AI-driven market analysis. It generates technical and fundamental insights, evaluates sentiment using Google Trends and Polymarket, and streams live interpretations through an intuitive web dashboard. In short, it empowers everyday investors to make data-informed decisions, backed by the same level of intelligence large financial institutions rely on.

How we built it

We built Velarith using a Next.js 14 + TypeScript frontend for a responsive and modern user interface, and a Flask + Python backend for analytical computations.

Frontend:

Developed using Next.js App Router, TailwindCSS, and Framer Motion for smooth UI transitions.

Integrated React Query for live data synchronization and Zustand for lightweight state management.

Backend:

Created with Flask 3.0, featuring blueprints for modular APIs.

Used pandas, NumPy, and ta libraries to compute over 30 technical indicators.

Integrated Claude Sonnet 4.5 via Anthropic’s SDK for contextual analysis.

Pulled live data from Finnhub, Yahoo Finance, and Polymarket APIs.

Deployment:

Frontend hosted on Vercel, backend on Render, both synchronized via GitHub CI/CD.

This structure allowed us to deliver streaming AI responses through Server-Sent Events (SSE), ensuring near real-time analysis updates.

Challenges we ran into

We encountered several major challenges:

Claude integration latency: Streaming AI responses required tuning SSE and token context to balance speed and completeness.

Data rate limits: Finnhub’s API throttling demanded caching layers and adaptive polling intervals.

CORS conflicts: Coordinating requests between Vercel (frontend) and Render (backend) required precise CORS origin management.

Type safety: Keeping a massive TypeScript and Python codebase in sync across hundreds of data fields was complex.

UI optimization: Balancing rich data visualizations with minimal load time required aggressive code-splitting and lazy loading.

Accomplishments that we're proud of

Building a full-stack, production-ready financial platform in a short hackathon window.

Achieving real-time AI market summaries using Claude Sonnet 4.5’s context window.

Designing a polished, mobile-optimized dashboard that feels professional yet intuitive.

Implementing multi-source data fusion, merging fundamentals, technicals, and sentiment.

Creating an open, modular backend that can easily be extended with new indicators or data providers.

What we learned

We learned that language models become powerful when grounded in quantitative data. Feeding structured financial metrics into Claude transformed raw numbers into insights understandable by all investors.

We also gained hands-on experience in:

Backend scalability and API orchestration,

Frontend performance optimization for streaming,

Collaborative full-stack development under intense deadlines,

Prompt engineering for domain-specific reasoning in finance.

What's next for Velarith

Enhanced AI reasoning: Multi-turn portfolio simulation and risk-reward optimization.

User personalization: Accounts, alerts, and saved dashboards.

Async backend: Migration from Flask to FastAPI for higher concurrency.

Expanded markets: Crypto, DeFi, and commodities integration.

Open developer API: Let others build on Velarith’s analytics engine.

Ultimately, our goal is to evolve Velarith into a comprehensive AI financial ecosystem, bringing institutional-level analytics to every trader.

Built With

  • anthropic-claude-4.5-api
  • finnhub
  • flask-3.0
  • framer-motion
  • github-ci/cd
  • google
  • next.js-14
  • numpy
  • pandas
  • polymarket
  • python-3.11
  • recharts
  • render
  • tailwind-css
  • typescript
  • vercel
  • yahoo-finance
Share this project:

Updates