Portfolio
Selected Work
Every project here was built to solve real problems through thoughtful design, clean engineering, and modern technologies. Each one reflects our approach to building fast, scalable, and user-focused digital experiences.
A full cinematic dark-UI redesign for an event planning and booking platform — built to feel as considered as the weddings and galas it books.
Uma Events needed a booking experience that looked as premium as the events it manages, not another template with a contact form bolted on.
We designed a full cinematic dark UI from scratch — an ink, gold and ivory palette paired with Bodoni Moda and Jost typography — then built an 8-table Supabase backend and a 9-module admin portal to run bookings, packages and media end to end.
Visual identity, front-end build, Supabase schema and the full admin portal, including a Cloudinary media pipeline for galleries and vendor assets.
A production booking platform with a real operations backend behind it — a tool the team runs the business on, not a brochure site.
A court-hearing management SaaS built for a working advocate, with WhatsApp automation doing the reminding so nothing slips through a crowded docket.
A practicing advocate was tracking hearing dates and outcomes by hand across dozens of active cases, with no automated way to remind clients or catch duplicate filings.
We built a hearing-management SaaS with automated WhatsApp reminders for hearing outcomes, bulk case import, duplicate-case detection, and multilingual messaging in English, Telugu and Hindi — deployed on a VPS behind PM2 for uptime.
End-to-end build: MongoDB schema, Express API, React front end, and the Baileys-based WhatsApp automation layer.
A real legal-tech product solving a practical, everyday problem for a working lawyer — live at vakilsummons.com.
An AI-powered stock forecasting platform, from data pipeline to mobile app — built to show range beyond business websites.
Retail investors on the NSE and BSE lacked an accessible tool that turned raw price data into a confident, explainable signal.
We trained a Temporal Fusion Transformer on NSE/BSE data, served it through a FastAPI backend, and wrapped it in a Flutter app with Zerodha Kite Connect integration and an OCR pipeline (Tesseract + OpenCV) that reads a user's existing portfolio straight from a screenshot.
Full stack: data pipeline, model training, FastAPI backend, Flutter mobile app and the OCR ingestion layer.
98.28% accuracy on NSE/BSE data with clear BUY/SELL/HOLD signals and confidence scoring — a complete AI product, not a notebook demo.
A couple more, built the same way.
Velvet Mochi
Serialised fiction needed a home that felt as immersive as the stories themselves — most publishing tools are built for text dumps, not for readers who return chapter after chapter.
We built Velvet Mochi from scratch: chapter-by-chapter publishing, a personal bookshelf, newsletter automation and comment moderation via Resend, and reading customization down to a personalized unsubscribe flow — on a Next.js and Supabase stack deployed on Vercel.
It's our own product, still live and running, with a returning reader base and engagement features doing retention work a small team can't do by hand.
Delhi Highway Sweets & Namkeen
Delhi Highway needed to sell online without losing the local, in-person trust that built the business — a generic storefront template wasn't going to cut it.
We shipped in phases: a WhatsApp Business integration for order confirmations, Razorpay for payments, an AI chatbot for product questions, and a location-based city selector so delivery and stock stay accurate per region.
Each phase went live independently, so the client was taking real online orders before the full build was even finished.