ShopSync:
Building India's AI-Powered
Retail Intelligence Platform

The Observation That Started It All

Walk into any kirana store in India and you'll notice something that hasn't changed in decades. Behind the counter, there's usually a worn notebook, a mental catalogue of what's in stock, and an owner who's running the entire operation on memory and intuition.

India has over 63 million of these small retail stores. Together they form the backbone of how over a billion people buy their daily essentials — milk, bread, eggs, snacks, cooking oil. Yet almost none of them have access to technology that helps them understand their own business. No inventory alerts. No sales trends. No sense of what to order before they run out.

While massive organised retail chains use sophisticated algorithms to track transaction velocity and forecast demand, 0% of these unorganised stores have access to AI-powered predictive tools. They run their businesses on intuition, memory, and pen-and-paper ledgers. When a store owner guesses wrong, they face out-of-stock items, lost sales, and wasted capital.

That observation, back in 2024, became the seed of ShopSync.

What ShopSync Is

ShopSync is an AI-powered retail management platform built specifically for kirana and small retail store owners in India. At its core, it does two things really well: it makes recording daily transactions completely effortless, and it turns those transactions into actionable intelligence.

The app runs on Android and was designed with one principle in mind — if it takes more than a few seconds, a store owner serving 50 customers a day won't use it.

By logging daily sales seamlessly through any of these methods, ShopSync transitions from a simple utility app into an intelligence engine. The transactions build a historical database, allowing the app to generate a personalised "Daily Pulse" dashboard. It analyses seasonal shifts — like the transition from peak summer to the monsoon season — and tells the shopkeeper exactly what to restock, how much, and when.

Meet Sunny

Sunny is ShopSync's AI assistant, and she's the heart of the platform.

Built using Google's Agent Development Kit (ADK) and powered by Gemini 2.5 Flash, Sunny is a multi-agent AI system with four specialised sub-agents working under her:

Sunny is voice-native. A retailer can tap a floating button, speak naturally, and Sunny records the transaction and updates the inventory without the owner touching anything else. She also responds in Hindi and regional languages, which matters enormously for the actual users of this product.

Ask Sunny what to stock for May and June, and she doesn't guess. She queries real sales data, identifies the summer spike in beverages and dairy, and gives you specific product recommendations backed by numbers. That's not a feature you'd expect in an app built for a kirana store. That's the point.

Building Lean and Scaling Smart

Building a product for the unorganised retail market requires a rigorous focus on feasibility and cost execution. The architecture has to be fast, reliable, and incredibly lean.

The Android app is built with Jetpack Compose and follows an offline-first architecture. All data is stored locally using Room database, so the app works even without internet. When connectivity is available, everything syncs automatically to Firebase Firestore in the background.

Sunny runs as a deployed service on Google Cloud Run. A separate MCP (Model Context Protocol) Toolbox service — also on Cloud Run — acts as the data pipeline, connecting Sunny to both Firestore (for live inventory reads) and Google BigQuery (for sales analytics and forecasting). This separation keeps the architecture clean and scalable.

The sales data powering the forecasting covers a full year of realistic Indian retail patterns — summer beverage spikes, festive season snack surges, monsoon comfort food trends — giving Sunny the context to make genuinely useful predictions.

Building this wasn't without its technical grit. Developing in this space means fighting through unexpected hurdles — like debugging hard network blocks where local ISPs fail to route Google Cloud traffic, requiring custom DNS bypasses just to keep the database syncing. It is a constant process of breaking things, fixing them, and hardening the code for real-world environments.

Forging the MVP

ShopSync started as a concept in 2024. The actual build began in earnest when I joined Google Cloud's GenAI Academy, organised by Hack2Skill. That program was instrumental in teaching me how to build and deploy functional AI agents integrated with Gemini — skills that became the foundation of Sunny.

The first version of the app was rough. Inventory data was hardcoded. The barcode scanner was simulated. The AI could barely answer a question. Over weeks of iteration — late nights, deployment failures, permission errors, and more than a few moments of genuine frustration — the product came together.

While the project didn't advance to the final 60 at FixForward, the true ROI of the competition was the acceleration in knowledge. Competing forces you to strip away the fluff, solve complex UX and backend challenges in days rather than months, and focus purely on execution. It proved that the technology works, the vision is solid, and the market need is urgent.

Early conversations about the concept with Arin Shrivastava, Arsh Shrivastav, and Ayush Sinha were invaluable — their input helped turn a rough idea into something worth building.

The Bigger Picture

ShopSync isn't just an inventory app. The long-term vision is more interesting.

Every store that uses ShopSync generates data. Every transaction is a signal. At scale — across thousands of stores, in dozens of cities — that becomes something no one has: a real-time, granular view of what India's unorganised retail sector is actually selling. By product. By city. By season.

That data has value well beyond the individual store owner. FMCG companies spend enormous resources trying to understand regional demand patterns. Supply chain distributors make costly mistakes because they can't see what's moving at the ground level. ShopSync, at scale, solves that.

ShopSync in Action

A short demo that provides a visual pitch of the app and showcases exactly how the ShopSync intelligence dashboard operates.

The Road Ahead

ShopSync is still very much a work in progress. The core product works — the barcode scanner, voice transactions, Sunny's forecasting, the Firestore sync — but there's a long list of things still to build. The mission to digitise 63 million stores is massive, but it starts by ensuring the core technology is absolutely flawless for just one transaction at a time.