Startup Spotter

• 

Inspiration

We were inspired by the idea that entrepreneurship should be accessible to everyone — not just people in tech hubs with venture funding. Millions of people in rural or underserved areas have the drive to start businesses but lack personalized, data-driven guidance. Traditional idea lists are generic, outdated, or disconnected from local market needs. We built Startup Spotter to bridge that gap — giving aspiring founders AI-generated business ideas based on real-world context like location, budget, and local demand.

What it does

Startup Spotter is an AI-powered startup discovery platform. Users enter simple inputs — like their location, budget, and interests — and receive personalized microbusiness ideas with summaries, key market insights, and recommended resources. It uses Google Cloud’s Vertex AI for natural language understanding and MongoDB Atlas Vector Search to match users with demographically relevant counties and proven business ideas. The platform also visualizes entrepreneurial “hotspots” using real census and economic data to help users identify strong market opportunities.

How we built it

We used a full-stack architecture with a modern, clean UI and seamless AI + data integration:

Frontend: React + TypeScript with Tailwind CSS, Framer Motion for animations, and a split-panel layout for search + results.

Backend: Node.js (Express) for routing and API logic, and Python for agent workflows and query handling.

AI Layer: Google Vertex AI (text-bison) powers the AI agent to interpret user input and generate recommendations.

Database: MongoDB Atlas stores startup data and census demographics, using Vector Search with embedded documents for similarity matching.

Session Handling: When a user starts a session, initial_setup.py fetches their context and stores a session ID; subsequent prompts are handled by query.py.

Visualization: We built a geo-based heatmap using public datasets to identify high-opportunity regions for microbusinesses.

Challenges we ran into

Vector Search Complexity: Integrating MongoDB’s vector search required preprocessing and embedding demographic JSON records — and tuning similarity thresholds to get meaningful matches.

Frontend-Backend Sync: Connecting the frontend search bar with Python-based agent workflows via Express and axios introduced async challenges and timing bugs.

AI Tuning: Vertex AI initially returned overly generic ideas. We iteratively refined prompts and injected context from vector matches to improve quality.

Hackathon Scope: Building a fully interactive, end-to-end AI experience — with a working heatmap, dynamic prompts, and data-backed outputs.

Accomplishments that we're proud of

Built a polished, animated UI with seamless AI interaction.

Learned about new state-of-the-art libraries such as MongoDB Atlas and Google ADK

Successfully integrated MongoDB vector search + Google Vertex AI, a non-trivial AI-data pipeline.

Generated high-quality, demographically relevant startup ideas with real-world utility.

Developed a personalized, engaging user experience — not just a proof of concept.

What we learned

How to use MongoDB’s Vector Search with custom embeddings and similarity scoring.

How to orchestrate multi-agent AI logic across Python scripts within a Node.js backend.

How to structure a React/TypeScript app for responsiveness and animation polish.

How to prompt and chain responses using Vertex AI to create a coherent experience.

Most importantly, how to deliver a well-scoped, functional, and impactful MVP under pressure.

What's next for Startup Spotter

We're excited to continue development! Here’s what’s on our roadmap:

Public Dataset Expansion: Ingesting larger, richer datasets to improve location-personalization and startup idea diversity.

User Profiles & Saved Ideas: Letting users save and revisit ideas, or get recurring recommendations.

Multi-Agent Pipeline: Implementing Google's ADK to support a more robust back-and-forth conversational agent.

Market Validation Tools: Adding demand forecasts, competitor analysis, and startup cost calculators.

Monetization: Exploring partnerships with microfinance platforms or grants for underrepresented founders.

Built With

• googleadk
• mcp
• mogodbatlas
• mongodb
• python
• react
• typescript
• vectorsearch
• vertexai