GramSell AI

• 

Inspiration

Every morning, millions of rural farmers and small sellers across Bangladesh, India, and South Africa wake up before dawn — not knowing if today's hard work will feed their family.

A tomato farmer in Sylhet has no way to know that today's price in his nearest market has dropped. A fish seller in Bihar cannot write a Facebook post because she never went to school. A handcraft maker in Johannesburg has a beautiful product but zero digital reach.

These are not lazy people. These are hardworking people trapped by a lack of information and tools. Unemployment and poverty in rural areas are not just economic problems — they are problems of access. Access to information. Access to markets. Access to financial tools.

GramSell AI was built to break that wall — to create real employment opportunities, real income growth, and real financial inclusion for the people who have been left behind by the digital economy.

What it does

GramSell AI is a full-stack AI-powered selling assistant designed for rural micro-entrepreneurs across three countries.

📸 Sell Tab — Digital Presence in Seconds The seller takes a photo of any product — vegetables, fish, poultry, handmade crafts, dairy. Gemini AI analyzes the image and instantly generates 5 completely unique product descriptions every time — each from a different emotional angle, never repeated. Ready-to-post Facebook, Instagram, and WhatsApp content is also generated with relevant hashtags. The seller's country determines the output language automatically — no manual selection needed. Payment terms and MFS number are automatically included in every post: bKash/Nagad/Rocket in Bangladesh, UPI/GPay/PhonePe in India, SnapScan/EFT in South Africa.The seller can also edit the generated post before sharing — adding personal touches while keeping payment and MFS details intact.

📊 Price + Margin Tab — Know Your Worth The seller enters their product name, nearest city, actual cost price, and desired margin (5%, 10%, 15%, or 20%). Google Search Grounding fetches real-time market prices. The AI calculates the recommended selling price, profit per KG, and the exact total price the buyer will see — so the seller never sells below cost and is never cheated by middlemen again.

🌦 Weather Tab — Smart Market Decision GPS auto-detects the seller's location and finds nearby markets, haats, and bazaars within 25km via Google Maps. Real weather data is fetched and the AI gives a clear verdict: GO TO MARKET, GO EARLY, or WAIT. It analyzes how today's weather impacts the specific product's price, warns about spoilage and transport risks, and recommends which nearby market is best to go to today.

💰 Finance Tab — Financial Inclusion for the Unbanked Emergency Fund Calculator: based on income, expenses, business type, and risk level, the AI calculates exactly how much money to keep as emergency savings, how many months it will take to reach that target, and how much to set aside each month — with local savings scheme recommendations (DPS/FDR in Bangladesh, PPF in India, Stokvel in South Africa).

Loan Eligibility Scorer: the seller enters income, age, loan amount, existing debts, and collateral. The AI gives a real credit score (0–100), approval chance, improvement tips, required documents, and recommended local banks — so a rural seller can understand their financial standing before approaching any bank.

How we built it

Frontend: Next.js 15 + React 19, custom dark green organic design with no UI framework. Deployed on Vercel.

Backend: Next.js API Routes as serverless functions. All Gemini AI calls, AWS DynamoDB operations, and Google Maps requests happen entirely server-side. No API keys are ever exposed to the browser.

Database: AWS DynamoDB with 5 production tables on On-demand capacity:

• gramsell_sellers — seller profiles with MFS payment details
• gramsell_products — AI-generated listings with full social content
• gramsell_prices — market price lookups with cost and margin data
• gramsell_weather — weather verdicts and nearby market data
• gramsell_finance — emergency fund calculations and loan scores

AI: Google Gemini 3.1 Flash Lite with multimodal image analysis and Google Search Grounding for real-time market and weather data.

Location: Google Maps APIs for GPS reverse geocoding and nearby market and bazar discovery within 25km.

Challenges we ran into

Designing for users with zero digital literacy was the hardest challenge. Every screen had to be self-explanatory — take a photo, tap a button, get results. No jargon, no complex menus.

Making AI output feel genuinely local was another challenge. Every Facebook post and WhatsApp message had to sound like it was written by a real person from that community — emotional, personal, and trustworthy — not like a robot.

Handling three countries, three currencies, three payment systems within a single clean architecture required careful design of the COUNTRY_LANG data model that drives the entire app behavior from one source of truth.

Accomplishments we're proud of

A woman selling nokshi katha in a village in Bangladesh can now photograph her quilt, get a beautiful post in her own language with her bKash number already included, know which nearby bazar is paying the best price today, check if rain is coming, calculate exactly how much emergency savings she needs and how many months to reach that target, and understand if she qualifies for a loan — all in one app, in under two minutes.

That is real empowerment. That is employment creation. That is what technology should do for the people who need it most.

What we learned

Real impact comes from understanding the real life of the person you are building for — their language, their fears, their daily decisions, their financial reality. AWS DynamoDB + Vercel + Gemini AI made it possible for a solo builder to ship a production-grade app that can scale to serve millions of rural sellers across three countries without re-architecting.

What's next for GramSell AI

• Expansion to Nigeria, Kenya, and Myanmar
• Partnership with NGOs and microfinance institutions for direct seller onboarding
• Group selling feature for farmer cooperatives

Built With

• aws-dynamodb
• google-gemini-3.1-flash-lite
• google-maps
• google-search-grounding
• next.js-15
• react-19
• vercel