Uddyog-CoPilot

Inspiration

Bangla-speaking entrepreneurs often lack access to structured business mentorship and strategic evaluation. Traditional advisory services are expensive and unavailable to early-stage founders. We built Uddyog-CoPilot to democratize startup execution by combining Multi-Agent AI debates, risk mapping, and seamless Web3 micropayments.

What it does

Uddyog-CoPilot is an AI-powered digital business assistant tailored for local entrepreneurs:

War Room (Multi-Agent Debate): Evaluates startup ideas via 3 distinct AI agents (Skeptic for risk analysis, Strategist for market growth, Operator for feasibility).
Assumption Map: Automatically categorizes claims into VALIDATED, UNVALIDATED, and NEEDS_INFO states.
Document Advisor (RAG): Allows pitch deck and PDF uploads to query business advisors trained on uploaded documents using Google Cloud Storage and pgvector.
x402 Micropayments: Integrates Web3 micropayment middleware to pay per deep AI evaluation without subscription models.
Pitch Session: Uses Gemini Live Multimodal API for real-time audio and camera feed pitch feedback in Bangla.

How we built it

Frontend: Next.js 15 (App Router), React 19, TypeScript, Tailwind CSS 4, and React Flow.
Backend & Database: Express 5, Prisma 6, PostgreSQL with pgvector extension.
AI Engine: Gemini Flash, Gemini Live Multimodal API, and GCS for RAG storage.
Devin Integration: Devin was used to audit backend Express routes, generate test cases, and optimize Prisma database queries.

Challenges we ran into

Balancing real-time token streaming and prompt synchronization across 3 concurrent AI debate agents.
Implementing low-latency audio/video feeds using Gemini Live Multimodal API in native Bangla context.
Structuring the x402 micropayment middleware to enforce instant payment challenges without breaking Next.js API route state.

Accomplishments that we're proud of

Built a fully localized, Bangla-first UX tailored specifically for South Asian startup ecosystems.
Successfully executed a working multi-agent debate framework with dynamic risk mapping.
Integrated x402 Web3 micropayments alongside standard ad slots for flexible monetization.

What's next for Uddyog-CoPilot

Automated Financial Projection Engine: Generate dynamic P&L and cash flow charts automatically.
Local Micro-financing Connectors: Connect validated business plans directly to local lenders and incubators.

Built With

agentic
ai
devin
express.js
gcs
gemini
next.js
pgvector
postgresql
prisma
protocol
rag
react
typescript
web3
x402

Submitted to

Agentic Hackathon

Created by

What I Worked On & Contributed:
Multi-Agent Architecture & War Room: Designed and built the core multi-agent AI system (Skeptic, Strategist, and Operator) using Gemini Flash to evaluate business risks and generate dynamic assumption maps.
Full-Stack & Payment Integration: Developed the Next.js 15 and Express backend pipeline, integrated Google Cloud Storage (GCS) for pitch deck RAG workflows, and implemented x402 micropayments middleware for AI feature gating.
Multimodal Pitch & Bangla UX: Integrated Gemini Live Multimodal for real-time audio/video pitch feedback and designed a fully localized, Bangla-first user interface.
What I Learned & What I'm Proud Of:
Successfully engineered a complex multi-agent AI debate framework that turns vague business ideas into structured, stress-tested execution plans.
Gained deep, hands-on experience combining Gemini's multimodal capabilities with Web3 micropayments (x402) and vector-based document search (RAG) on Google Cloud.
Delivered a tailored, high-impact AI co-pilot built specifically for Bangla-speaking entrepreneurs and early-stage startups.

MD. ABDUR RAHMAN HRIDOY
Tech enthusiast and college student. Eager to learn, collaborate, and build innovative projects at hackathons.
MD Riad Hossain