RAG · SheLeads 2.0 · LeadHer track
Matching women to the schemes built for them.
An AI-powered scheme matching engine from Team CodeCorps. It connects Indian women to the central and state financial programmes that fit them, using a retrieval-augmented pipeline over local LLMs, a vector database and a document store, then explains eligibility in plain language.
WebGL · move your cursor
0
Pipeline stages
Top 0
Schemes retrieved per query
RS256 JWT
API auth
SDG 5
Aligned with
// highlights
Whatmakesittick
Profile in, schemes out
Matches on state, age, income, occupation and purpose (business, education, savings) pulled from a plain chat message.
Eligibility in plain language
An LLM turns raw scheme matches into customised eligibility explanations and recommendations anyone can follow.
Built for the hardest-to-reach
Aimed at women entrepreneurs, self-help group members and rural applicants who struggle with fragmented portals and complex criteria.
Local or remote LLMs
Ollama runs on the same machine or on a separate GPU box on the network, so a light laptop can still serve a heavy model.
A real data pipeline
Scripts scrape, clean and vectorize scheme data so natural-language queries can run over unstructured documents.
Locked-down API
Every /api route checks the user's Clerk JWT signature against Clerk's JWKS before doing any work.
// architecture
Howit'sbuilt
- 01React 19 + ViteClerk auth · Framer Motion · R3F
- 02FastAPIAsync · Pydantic · Motor
- 03Ollama · Qdrant · MongoDBLLM · vectors · scheme documents
step 01
Extract
FastAPI asks Ollama to turn "I'm a 24-year-old female farmer in Karnataka" into a structured JSON profile.
step 02
Retrieve
The profile and query are searched against Qdrant to find the top 5 semantically relevant schemes.
step 03
Enrich
Scheme IDs fetch full text and eligibility criteria from MongoDB's detailed_schemes collection.
step 04
Generate
Vector results, deep details and the profile go back to Ollama for a helpful, accurate reply.
// under the hood
Thedetails
{
"age": 24,
"gender": "female",
"occupation": "farmer",
"state": "Karnataka",
"income": null
}// stack
Builtwith
- React 19
- Vite
- Tailwind
- Framer Motion
- React Three Fiber
- Clerk
- FastAPI
- Ollama
- Qdrant
- MongoDB
- Pydantic