TaxMitra

AI-powered CA assistant for India's 63M small businesses. Get instant GST, ITR & TDS guidance in Hindi & English for just ₹99 = ~$1.20. Built with Google Gemini 2.5 Flash. 9 days. 1 Developer(Me).

Comment

Inspiration

India has 63 million small businesses. A kirana store owner in Jaipur, a freelance designer in Mumbai, a textile trader in Surat — they all face the same problem: complex GST filings, ITR deadlines, TDS compliance. The solution? A CA charging ₹5,000–₹20,000/month(₹5,000 = ~$60 | ₹20,000 = ~$240 ). Most cannot afford it.

I discovered this hackathon with only 9 days remaining. While others had 82 days, I had 9. One question drove me: what can one person build with AI in 9 days that solves a real problem for millions?

TaxMitra is the answer. Built in 3 days. Submitted before 7 Days to deadline.

What it does

TaxMitra is NOT a general AI chatbot. It is a specialized AI Chartered Accountant — purpose-built exclusively for Indian tax law.

Why TaxMitra is different from ChatGPT, Gemini, or other AI tools:

Feature ChatGPT/Gemini Other Tax AI TaxMitra
Indian tax specialization Generic Partial 100% focused
Hindi + English English first English only Both automatically
Exact section numbers Sometimes Rarely Always
Price ₹1,700/month ($20/month) ₹500-2,000/month (₹5,000 = ~$60 ₹20,000 = ~$240
Target user Tech-savvy CA professionals Small business owners
Built for India No Partial Yes

The AI handles everything autonomously:

  • GST registration, GSTR-1, GSTR-3B filing with exact section numbers
  • ITR filing (ITR-1 to ITR-4) with deadlines
  • TDS calculation and quarterly return guidance
  • Tax notice interpretation and response drafting
  • Tax saving strategies — Section 80C, 80D, HRA
  • Compliance calendar — never miss a deadline
  • Responds in Hindi AND English automatically

Every response is generated by Google Gemini 2.5 Flash. No human reviews or edits any answer. The AI IS the CA.

How we built it

Built in 3 days on Google Cloud:

Component Technology
AI Brain Google Gemini 2.5 Flash (Vertex AI)
Backend Python Flask
Deployment Google Cloud Run
Database Google Cloud Firestore
Payments Razorpay (Live)
Repository GitHub (Public)

AI Operations Evidence: Every user interaction is permanently logged in Google Firestore. The /logs endpoint provides real-time evidence of AI running in production continuously — timestamps, queries, responses, all stored autonomously without human intervention.

Payment Infrastructure: Live Razorpay integration processes real payments. The AI handles the product. Razorpay handles the money. Zero human involvement end-to-end.

Challenges we ran into

  • 3 days to build everything: Ruthless prioritization — product, payments, deployment, security all in 72 hours
  • Vertex AI integration: Configuring Gemini 2.5 Flash on Cloud Run with proper service account authentication
  • Bilingual AI responses: Prompting Gemini to consistently respond in both Hindi and English with expert-level accuracy including exact Indian tax section numbers
  • Security incident: GitGuardian detected exposed Razorpay keys within hours of pushing to GitHub — immediately rotated keys, moved to environment variables on Cloud Run.
  • Payment UX: Integrating Razorpay checkout flow seamlessly without breaking JavaScript in the Flask HTML template

Accomplishments that we're proud of

**The 3-Day Build

Day Milestone
Day 1 Product built, deployed on Google Cloud Run
Day 2 Live Razorpay payments integrated, security hardened
Day 3 First paying customers, Firestore logging live ,Monitoring, refinements
Day 3-4 Demo video recorded, submission complete

Revenue: ₹598 (~$7 USD) from first paying customers within 72 hours of payment integration going live.

The velocity argument:

  • 3 days to build → paying customers within 48 hours of launch
  • 82 days (full hackathon) → exponentially more customers and revenue
  • With marketing budget → market of 63 million is addressable

The AI runs 100% of operations. Scaling from 2 to 20,000 customers requires zero additional headcount. Google Cloud Run scales automatically. Gemini handles every query. This is a truly AI-native business.

What we learned

  • Gemini 2.5 Flash delivers expert-level domain knowledge rivaling human professionals when properly prompted
  • Speed of execution — 3 days of focused building beats months of slow planning
  • India's tax compliance gap is a $50 billion opportunity hiding in plain sight
  • AI-native businesses can be built, deployed, and monetized by a single founder in days not years
  • Security matters — real production systems need proper secret management from day one
  • Google Cloud provides enterprise-grade infrastructure accessible to solo founders instantly

What's next for TaxMitra

The 3-day experiment proved three things:

1. The payment infrastructure works — live Razorpay transactions processed successfully, proving the monetization model is technically sound and ready to scale

2. AI runs the business autonomously — zero human intervention from query to answer. Gemini 2.5 Flash handles every tax question with expert-level accuracy 24/7

3. Google Cloud scales instantly — same infrastructure that serves 2 users today serves 2 million tomorrow with zero changes

Why this matters beyond the hackathon:

The competition is not ChatGPT or Gemini. The competition is the ₹10,000/month($120/month) CA. TaxMitra wins that competition at ₹99($1.20).

  • India: 63 million small businesses — addressable immediately
  • Global: 500 million+ small businesses worldwide face identical problems
  • Same model, different tax systems — GST in India, VAT in Middle East, similar compliance needs globally

The infrastructure is built. The AI is running. The payments work.

3 days. 1 Developer. India first. Global next. 🌍

Built With

Share this project:

Updates