Loading
AI / SaaS
StartupSage
An advanced, context-aware AI co-pilot and Agentic RAG platform for Indian founders. Navigates government policies, schemes, taxation, DPIIT recognition, and legal compliance with zero hallucinations and official citations.
React 19
FastAPI
LangGraph
Qdrant
Gemini API
Tailwind CSS
The Overview
StartupSage is an advanced, context-aware AI advisor built specifically for Indian founders navigating the complex ecosystem of government policies, DPIIT recognition, taxation, MSME guidelines, and funding schemes. Powered by an Agentic RAG pipeline using LangGraph, FastAPI, and Qdrant, it autonomously routes queries, rewrites ambiguous prompts, and verifies groundedness against official Indian Government Gazettes and Acts to eliminate hallucinations. The platform features personalized startup profiling (filtering advice by stage, industry, and state), document attachment analysis for custom legal files, real-time voice dictation with Web Speech API, and a striking neo-brutalist dark interface with smooth Lenis scrolling.
Key Features
Agentic RAG workflow powered by LangGraph with autonomous query rewriting & verification
Context-aware advisory filtering policies by startup industry, stage, and state
Direct PDF/TXT document upload analysis for instant legal contract Q&A
Native Web Speech API integration for real-time voice-based querying and TTS read-aloud
Zero-hallucination responses with verified official government citations and source links
Bold neo-brutalist UI with high contrast, vibrant accents, and smooth Lenis scrolling
The Journey
"Designing an enterprise-grade ingestion and chunking pipeline using PyMuPDF to preserve complex tabular and hierarchical structures from official Indian government gazettes was a primary challenge. Developing an Agentic RAG supervisor with LangGraph that accurately classifies intents, rewrites multilingual/Hinglish queries, and reranks chunks using cross-encoders without introducing latency required extensive optimization. Implementing token-by-token async response streaming in FastAPI alongside background model pre-warming ensured seamless real-time interaction on the neo-brutalist frontend."
Showcase Portfolio
Next Exploration
Let's build something epic→