MarketMinder Ai

• 

  The settings page, where the general setup for the shop is made

• 

  A section for the general overview of the Business

• 

  Ai chatbot, an Ai tool for conducting market research , processing receipts, or inputting stock or sales either using text or voice notes

• 

  A section for showing a general overview of the business reports, for compliance purposes

• 

  A section for showing the general business health, inventory velocity and generation of market strategy

• 

  Intelligence Guide section, showing the price anatomy of the Business, indicating the floor price, the benchmark price and the realized sale

• 

  A section showing the smallest unit normalization

• 

  The strategy and the health score of the business

• 

  Sales Terminal

• 

  Business Health

Inspiration

The concept for MarketMinder AI emerged from observing Nairobi traders in Gikomba and Wakulima struggling with manual records and loan barriers, inspired by FSD Kenya reports on the 80% informal workforce. Gemini 3's multimodal AI sparked a conversational tool to digitize processes, empowering 2.8 million+ traders to break poverty cycles through accessible fintech.

What it does

MarketMinder AI is a conversational ERP for Kenyan traders, leveraging Gemini 3 to parse voice, text, and receipts for real-time stock tracking, cashflow forecasts, and credit scoring. It generates audit-ready reports to reduce predatory lending and enable formal loans like KCB-M-Pesa.

How it was built

The app was built with React 19 and Tailwind CSS for the UI (dashboard, chat, ledger). Gemini 3 (flash for parsing, pro for insights) handles multimodal inputs via API calls, with React Hooks/LocalStorage for persistence. Offline caching and mock M-Pesa integrations were added; testing incorporated trader scenarios for Swahili support.

Challenges faced

Accuracy for noisy inputs (blurred receipts, Sheng) required prompt tweaks and testing. Mobile optimization for poor networks necessitated redesigns. Compliance with Kenya's Data Protection Act involved adding consents and PIN locks amid time constraints, with prioritization of core features.

Accomplishments I'm proud of

95%+ multimodal parsing accuracy was achieved, along with localization for Kenyan dialects and a tool impacting 2.8M traders by enabling credit access. Gemini was integrated seamlessly for voice-to-ledger, creating a compliant app in a short hackathon window.

What was learned

Development highlighted AI's potential in low-literacy markets via prompt engineering for cultural nuances and Gemini optimization. Insights from FSD studies revealed informal vulnerabilities, emphasizing inclusive design, React state efficiency, and fintech privacy.

What's next for MarketMinder AI

Expansion to real M-Pesa/bank APIs for live loans, addition of chamas support, and scaling pan-Africa are planned. Trader feedback will inform price predictions, partnerships with FSD Kenya, and funding pursuits to reach 1M+ users for inclusive growth.

Built With

• gemini-2.5-flash-preview-tts-for-voice)
• gemini-3-pro-preview-for-analytics
• google-gemini-api-(@google/genai-with-gemini-3-flash-preview-for-parsing
• lucide-react-(icons)
• react-19
• react-hooks-(usememo
• react-router-dom-(hashrouter)
• recharts-(charts)
• tailwind-css
• usecallback)-with-localstorage