Backend-leaning Full Stack Developer focused on building production-ready applications, intelligent backend systems, REST APIs, authentication flows, database-driven products, and AI-powered experiences.

I enjoy taking projects from architecture → implementation → deployment → real-world debugging.

AI-powered financial intelligence assistant built for Telegram. Atlas combines persistent memory, multi-source financial data, SEC filings, personalized watchlists, contextual AI reasoning, and proactive market monitoring.

Highlights

• 🧠 Persistent short-term and long-term memory using PostgreSQL + Prisma
• ⚡ Intent-aware routing for financial requests
• 📈 Finnhub market data with Yahoo Finance fallback
• 📰 Marketaux financial news intelligence
• 🏛️ SEC EDGAR regulatory filing research
• 🎯 Personalized analysis using user memory and watchlists
• 🔔 Background watchlist monitoring with ±3% significance threshold
• 🛡️ Database-backed alert cooldown and deduplication
• 🤖 Google Gemini reasoning layer
• 🚀 Deployed Telegram bot and backend

🔗 Live Bot: https://t.me/AtlasFinancialAssistantBot
🔗 Live Backend: https://atlas-ai-financial-assistant-zrtk.onrender.com
🔗 Repository: https://github.com/jhasaurav97/atlas-ai-financial-assistant
🔗 Demo: https://www.loom.com/share/94998c29ad4445be9eaf87e8a120d294

Note: The live backend uses Render's free tier and may sleep when inactive. Open the backend URL once before testing the Telegram bot so the service can wake up.

A framework-agnostic headless media platform built as a TypeScript monorepo with strict package boundaries between the core SDK, React wrapper, headless UI library, and Vite web app.

Highlights

• TypeScript-first, framework-agnostic media core SDK
• React wrapper and headless UI component architecture
• Strict npm workspace dependency boundaries
• Paginated search and video-search APIs
• In-memory request de-duplication cache
• Headless hooks and event-driven interactions
• Clean separation between business logic, wrappers, and UI
• Vite web application consuming the SDK packages

🔗 Live App: https://fotoowl-headless-media-sdk-web.vercel.app/
🔗 Repository: https://github.com/jhasaurav97/fotoowl-headless-media-sdk

An AI-powered full-stack web application built to turn user input into meaningful insights through a responsive React interface and a production-style backend. The project focuses on real-world engineering concerns including authentication, validation, database persistence, secure API communication, deployment, and failure handling.

Highlights

• 🤖 AI-powered insight generation through backend services
• 🔐 JWT-based authentication with protected routes
• 🗄️ PostgreSQL database with Prisma ORM for schema management and persistence
• ⚡ RESTful backend APIs built with Node.js + Express.js
• 🎨 Responsive React UI with Tailwind CSS and light/dark theme support
• 🛡️ Input validation, error handling, and secure API communication
• 🐳 Dockerized backend and production-style deployment workflow
• 🚀 Frontend deployed on Vercel and backend deployed on Render
• 🔧 Resolved real deployment issues including Vercel SPA refresh/404 handling, CORS, and Prisma/Docker compatibility

🔗 Live App: https://insightforge-nine.vercel.app
🔗 Repository: https://github.com/jhasaurav97/insightforge

• Production-style REST APIs and backend services
• Authentication and authorization systems
• PostgreSQL/MongoDB data models and persistence layers
• AI-powered applications and API orchestration
• Full-stack applications with React and Node.js
• TypeScript SDK and headless UI architectures
• Dockerized and deployed applications
• Background processing and resilient integrations
• Debugging of real deployment and integration issues

• LinkedIn: https://www.linkedin.com/in/jha-saurav97
• GitHub: https://github.com/jhasaurav97
• Email: jhasaurav593@gmail.com