AI-powered web app that helps Indian citizens discover every Central & State government welfare scheme they're eligible for — in English, Hindi, and Marathi.

Instead of manually digging through dozens of scattered government portals, users either chat naturally ("I'm a 24-year-old farmer from Maharashtra earning ₹80k/year") or fill a quick profile form (age, income, caste, gender, profession, state), and the app returns matched schemes — PM Kisan, Ladki Bahin Yojana, Ayushman Bharat, PMAY, scholarships, pensions, etc. — with an AI-generated, plain-language explanation of why each one applies to them.

Architecture

React 19 + TS Frontend (Hugging Face Spaces / Render, Nginx+Docker)
↓ HTTPS
Cloudflare Worker (edge gateway)
↓
ASP.NET Core 8 Web API (Railway/Render, Docker)
↓
PostgreSQL (Neon) — 4,600+ schemes aggregated from 5 govt data sources

Key features

- Conversational AI matching — Claude (Anthropic) extracts a structured user profile from free-text chat, asks follow-up questions if data is missing, then explains matched schemes conversationally.
- Multi-source scheme aggregation — scheduled background jobs scrape/sync myscheme.gov.in (official API), india.gov.in, data.gov.in, and other sources, deduping and upserting into Postgres automatically.
- Multilingual (EN/HI/MR) — full UI and scheme content translation via Claude, with a free-tier fallback and caching to control API cost.
- Live fallback search — if the local DB has zero hits, it queries MyScheme.gov.in live so search results are never empty.
- Smart search & filters — fuzzy full-text search (Fuse.js) with autocomplete, category/state/profession/income filters.
- SEO-optimized landing pages — dedicated pages per major scheme (PM Kisan, Ladki Bahin, PMAY, student scholarships, MahaDBT) for organic discoverability.
- Bookmarks & WhatsApp sharing for saving/sharing matched schemes.
- On-demand detail scraping — full scheme details are lazily scraped and cached on first view rather than upfront for all 4,600+ schemes.

Tech stack

┌──────────┬───────────────────────────────────────────────────────────────────────────────┐
│ Layer │ Tech │
├──────────┼───────────────────────────────────────────────────────────────────────────────┤
│ Frontend │ React 19, TypeScript, Vite 8, Axios, Fuse.js │
├──────────┼───────────────────────────────────────────────────────────────────────────────┤
│ Backend │ ASP.NET Core 8 (C#), Entity Framework Core 9 │
├──────────┼───────────────────────────────────────────────────────────────────────────────┤
│ Database │ PostgreSQL (Neon) │
├──────────┼───────────────────────────────────────────────────────────────────────────────┤
│ AI │ Claude API (Anthropic) — chat extraction, explanations, translation │
├──────────┼───────────────────────────────────────────────────────────────────────────────┤
│ Scraping │ Selenium, HtmlAgilityPack │
├──────────┼───────────────────────────────────────────────────────────────────────────────┤
│ Infra │ Docker, Cloudflare Worker (edge gateway), Railway/Render, Hugging Face Spaces │
└──────────┴───────────────────────────────────────────────────────────────────────────────┘

What this demonstrates (for a Freelancer portfolio pitch)

- End-to-end full-stack delivery: React/TS frontend + .NET backend + Postgres, all containerized and deployed across 3 different hosting platforms.
- Practical LLM integration beyond a chatbot wrapper — structured extraction, multi-turn conversation state, cost-aware caching, multilingual generation.
- Real-world data engineering: multi-source scraping, scheduling, dedup/upsert pipelines against live, messy government data.
- Production concerns handled: CORS across origins, auto DB migrations on boot, health checks, edge gateway, SEO.