Engineering Real Business Impact

The client relied on slow legacy desktop software that couldn't sync live gold prices, manage Karigar inventories, or automatically file GST. Gold items > 2g had to comply with strict HUID hallmarking regulations. Manual entry led to calculation errors, bullion hedging delays, and compliance risks.

Designed and shipped a web/mobile ERP powered by a 20-agent system. Integrated IndexedDB POS cache to prevent network failures, automated gold/silver rate scrapers, tenant-isolated rate limiting, and BIS hallmark API integrations. Built a WhatsApp bot for client notifications.

Reduced billing times from 10 minutes to under 3 minutes (70% reduction). Enabled 100% automated hallmarking compliance and eliminated inventory leakage across workshop runs.

Soni Jewellery needed a premium digital showroom capable of scaling during wedding seasons. The biggest hurdle was pricing: jewellery prices needed to change dynamically with global bullion rates, while matching each online item with its precise offline HUID code.

Built a custom Next.js 16 storefront utilizing React 19 concurrent features. Created a serverless live commodity pricing engine updating inventory values automatically. Integrated Razorpay Checkout with webhook parity verification and dynamic cart state management.

Increased checkout conversion rate by 35%. Maintained 99/100 Lighthouse performance, securing stable operations during peak shopping traffic.

Rural patients in Western Rajasthan had to travel over 100 km for basic consults. Existing video services were too heavy for weak 3G/4G connections, and remote healthcare centres lacked automated prescription logs.

Engineered a lightweight React Native app with Twilio video compression. Created a FastAPI backend utilizing low-bandwidth fallbacks (JSON prescriptions over SMS) and connected clinics with regional medicine dispatch channels.

Served 10K+ rural patients within 6 months. Reduced travel costs by 85% on average and expedited emergency consultations to under 30 minutes.

Dryland farmers in Jodhpur suffered crop damage from late-detected pest infestations and lacked transparent mandi market rates, often forcing them to sell yields below production cost.

Built a bilingual Flutter application featuring an on-device PyTorch Mobile image classifier for crop health. Connected it to mandi price scrapers and logistics partners for fair negotiation and transport routing.

Saved crop losses estimated at ₹2Cr+ across 25,000 farmers. Farmers achieved a 22% average increase in net profit margins due to better timing and reduced middleman costs.